A method for treating tobacco wastewater

Abstract

This application provides a method for treating tobacco wastewater, comprising: Step S1: acquiring standard tobacco production plan data and standard tobacco wastewater full-spectrum data; Step S2: processing the tobacco production plan data according to a material balance model to obtain predicted influent water quality parameters, including predicted influent chemical oxygen demand (COD) concentration; Step S3: processing the tobacco wastewater full-spectrum data according to a feature extraction model to obtain real-time influent component feature vectors; Step S4: determining the reagent dosage parameters for the tobacco wastewater based on the predicted influent water quality parameters and the real-time influent component feature vectors. This invention, by acquiring production plan data from the upstream tobacco production line and simultaneously collecting full-spectrum data of tobacco wastewater, can obtain forward-looking information on water quality changes before the wastewater enters the treatment unit, completely overcoming the technical limitations of traditional wastewater treatment methods that can only provide delayed responses.

Description

Technical Field

[0001] This application belongs to the field of tobacco wastewater treatment, and specifically relates to a method for treating tobacco wastewater. Background Technology

[0002] Tobacco sheet production employs a papermaking process, using tobacco stems, broken tobacco leaves, and tobacco dust as main raw materials. Through core processes such as extraction, pulping, and forming, reconstituted tobacco leaves are produced. The wastewater generated during this production process is characterized by high pollutant concentrations, large fluctuations in water quality, and complex composition. The wastewater contains various pollutants, including nicotine, lignin, pectin, proteins, and polysaccharides, with its chemical oxygen demand (COD) concentration typically ranging from 8000 mg / L to 12000 mg / L. Due to the diverse sources of tobacco raw materials and frequent adjustments to the production formula, the wastewater quality exhibits significant fluctuations in both time and space, posing a considerable challenge to the stable and efficient operation of subsequent wastewater treatment systems. Currently, the tobacco sheet wastewater treatment industry generally adopts a combined process of "physicochemical pretreatment + anaerobic biological treatment + aerobic biological treatment + advanced treatment." The appropriate dosage of chemicals in the physicochemical pretreatment stage directly affects the stable operation of the subsequent biological treatment system and has a crucial impact on reducing overall treatment costs.

[0003] In existing technologies, various automatic control systems for wastewater treatment chemical dosing have emerged, attempting to address the problem of insufficient precision in chemical dosing. For example, Chinese invention patent CN110862188B discloses a precise chemical phosphorus removal dosing system and control method. This system consists of a feedforward control component and a feedback control component. The feedforward control component calculates the theoretical feedforward dosage based on the influent orthophosphate load, while the feedback control component adjusts the dosage in real time based on the deviation between the effluent orthophosphate concentration and the preset target value, thereby achieving precise control of phosphorus removal efficiency. Furthermore, some existing technologies utilize real-time monitoring of raw water flow and turbidity parameters for feedforward compensation, and combine this with flocculent image recognition technology to establish a feedback optimization model, forming a closed-loop control of the chemical dosing amount, further improving the timeliness of dosing control.

[0004] However, although the aforementioned existing technologies have achieved automatic dosing control based on influent water quality parameters, their control decisions all rely on real-time detection data after the wastewater enters the treatment unit. Essentially, they are either a delayed response or a simple feedforward adjustment based only on the current water quality. They cannot predict sudden changes in water quality caused by upstream production fluctuations and are difficult to meet the treatment needs of tobacco sheet wastewater with large water quality fluctuations.

[0005] The core limitation of existing chemical dosing control technology lies in the fact that all control decisions are based on the premise that "wastewater has been generated and entered the treatment system," lacking the ability to anticipate and coordinate upstream production conditions. When the wastewater quality in the upstream tobacco sheet production line changes abruptly due to operations such as formula adjustments, equipment cleaning, or process switching, the control system can only initiate adjustments to the dosing strategy after detecting the change in water quality. This lag often leads to a short-term decrease in treatment effectiveness and the risk of effluent exceeding standards. It can also result in waste of chemicals due to untimely or excessive dosing, increasing treatment costs.

[0006] More importantly, the nicotine and other components in tobacco sheet wastewater have high resource recovery value. However, the existing control strategies only focus on pollutant removal and fail to dynamically switch treatment modes according to the differences in the composition of the influent. As a result, valuable components such as nicotine are removed along with other pollutants, which increases the operating burden of the wastewater treatment system and causes potential waste of resources.

[0007] Therefore, how to make advance predictions on changes in the water quality of tobacco sheet wastewater and dynamically match differentiated reagent dosing strategies based on the characteristics of wastewater components, thereby overcoming the lag and single-target limitations of existing technologies, has become a technical problem that urgently needs to be solved in this field. Summary of the Invention

[0008] In view of this, the purpose of this application is to provide a method for treating tobacco wastewater to solve the above problems.

[0009] To solve the above-mentioned technical problems, this application adopts the following technical solution: This application provides a method for treating tobacco wastewater, comprising: step S1: acquiring standard tobacco production plan data and standard tobacco wastewater full-spectrum data; step S2: processing the tobacco production plan data according to a material balance model to obtain predicted influent water quality parameters, including predicted influent chemical oxygen demand (COD) concentration; step S3: processing the tobacco wastewater full-spectrum data according to a feature extraction model to obtain real-time influent component feature vectors; and step S4: determining the reagent dosing parameters for the tobacco wastewater based on the predicted influent water quality parameters and the real-time influent component feature vectors.

[0010] Furthermore, step S1 includes: step A1: obtaining tobacco production plan data within a preset time window; step A2: performing a first preprocessing on the tobacco production plan data to obtain standard tobacco production plan data, wherein the first preprocessing includes at least one of normalization, dimensionless processing, outlier removal, and data format unification.

[0011] Furthermore, step S1 includes: step A3: acquiring full-spectral data of tobacco wastewater that has not yet flowed into wastewater treatment; step A4: performing a second preprocessing on the full-spectral data of tobacco wastewater to obtain standard full-spectral data of tobacco wastewater, wherein the second preprocessing includes at least one of smoothing and noise reduction and baseline correction.

[0012] Furthermore, the standard tobacco production plan data includes the production output of the production formula, the pollution coefficient of the production formula, and the influent flow rate. The calculation formulas for the material balance model include:

[0013] in, In the future Predicted influent chemical oxygen demand concentration after a certain time. Indicates the current time Time The yield of this production formula Indicates the first The pollution coefficient corresponding to the production formula In the future Predicted inflow rate after time, This indicates the total number of production formulas.

[0014] Furthermore, the real-time influent component feature vector includes nicotine feature values, lignin feature values, pectin feature values, and protein feature values. The calculation formula for the feature extraction model includes:

[0015] in, This represents the feature vector of real-time influent components. This represents the full spectral data of standard tobacco wastewater. This represents the load matrix obtained through training with historical spectral data.

[0016] Furthermore, the chemical dosing parameters include the type of chemical dosing and the dosing rate. Step S4 includes: Step S41: matching the real-time influent component feature vector with the preset treatment target mode to determine the type of chemical dosing; Step S42: determining the dosing rate of chemical dosing based on the predicted influent flow rate, the predicted influent water quality parameters, and the type of chemical dosing.

[0017] Further, step S4 includes: Step S43: Based on the production formula code of the standard tobacco production plan data, call the pre-stored formula-component feature spectrum library to generate a predicted component feature vector; Step S44: Calculate the deviation data between the real-time influent component feature vector and the predicted component feature vector, and obtain the deviation correction coefficient based on the deviation data; Step S45: Input the deviation correction coefficient into the incremental proportional-integral-derivative control algorithm to adjust the type of pesticide and / or the pesticide dosing rate.

[0018] Furthermore, the incremental proportional-integral-derivative control algorithm is as follows:

[0019] in, Indicates the current time Correction amount for drug dosing rate. Indicates the current time The deviation between the real-time influent component feature vector and the predicted component feature vector. This represents the deviation value at the previous moment. This represents the deviation value between the two previous moments. Represents the proportionality coefficient. Represents the integral coefficient. This represents the differential coefficient.

[0020] Furthermore, the preset treatment target modes include a pollutant removal mode, under which... The agent being administered is polyaluminum chloride (PAC). The mathematical expression for the predictive dose-response model of PAC is as follows:

[0021] in, The predicted dosing rate of polyaluminum chloride is expressed in L / hour. To predict the inflow rate, the unit is cubic meters per hour. To predict the influent chemical oxygen demand concentration, the unit is mg per liter; The agent being administered is polyacrylamide. The mathematical expression for the predictive dose-response model of polyacrylamide is:

[0022] in, The predicted dosing rate of polyacrylamide is expressed in L / hour. The inflow rate is predicted in cubic meters per hour.

[0023] Furthermore, the preset treatment target modes include a nicotine recovery mode. Under the nicotine recovery mode, The agent used is a calcium-magnesium-aluminum composite flocculant. The mathematical expression for the predictive dose-response model of the calcium-magnesium-aluminum composite flocculant is as follows:

[0024] in, The predicted dosing rate of calcium-magnesium-aluminum composite flocculant is expressed in L / hour. To predict the inflow rate, the unit is cubic meters per hour. The nicotine concentration corresponding to the real-time influent component feature vector is expressed in mg per liter. The agent administered is a micro-sand suspension. The mathematical expression for the predictive dose-response model of the micro-sand suspension is as follows:

[0025] in, The predicted dosing rate of the microsand suspension is expressed in L / hour. The inflow rate is predicted in cubic meters per hour.

[0026] As can be seen from the above technical solution, the advantages and positive effects of the tobacco wastewater treatment method proposed in this application are as follows: I. This invention acquires production plan data from upstream tobacco production lines and simultaneously collects full-spectrum data of tobacco wastewater. It analyzes and processes the production plan data using a material balance prediction model and performs feature extraction on the full-spectrum wastewater data to generate predicted influent water quality parameters and real-time influent component feature vectors. This allows for the acquisition of forward-looking information on water quality changes before the wastewater enters the treatment unit, completely overcoming the limitations of traditional dosing systems that can only provide delayed responses. This enables proactive control and early execution of chemical dosing, effectively addressing sudden water quality changes caused by upstream production fluctuations, avoiding short-term declines in treatment efficiency and exceeding effluent quality standards. It also reduces waste caused by untimely or excessive chemical dosing, ensuring the long-term stable operation of the wastewater treatment system.

[0027] II. This invention matches the characteristics of influent components with a preset treatment target mode, retrieves a suitable reagent formulation, and combines this with a real-time correction unit to complete closed-loop correction and continuous optimization of prediction deviations. It can dynamically switch differentiated reagent dosing strategies based on the specific differences in influent components. This method ensures that pollutant removal efficiency meets preset standards, enables the recovery and utilization of valuable components in wastewater (such as nicotine), significantly improves the control precision and operational stability of reagent dosing, effectively reduces the overall operating cost of wastewater treatment, and achieves the dual benefits of environmental governance and resource recovery. Attached Figure Description

[0028] The above description of this application and the following detailed embodiments will be better understood when read in conjunction with the accompanying drawings. It should be noted that the drawings are merely examples of the claimed technical solutions.

[0029] Figure 1 This is a diagram showing the overall architecture and data flow of the tobacco wastewater treatment system; Figure 2 This is the internal logic diagram of the material balance prediction module of the tobacco wastewater treatment system. Figure 3This is a flowchart of the dual-mode driven intelligent decision engine control process for a tobacco wastewater treatment system. Figure 4 This is a control diagram of a tobacco wastewater treatment system under unsteady-state conditions. Figure 5 This is a flowchart of the iterative process for the effect verification and model self-learning module of the tobacco wastewater treatment system; Figure 6 This is a flowchart of a tobacco wastewater treatment method.

[0030] The reference numerals in the attached figures are explained as follows: Manufacturing Execution System Interface Unit: 20; Water ingress spectral fingerprint online monitoring unit: 21; Material balance prediction module: 22; Dual-mode driven intelligent decision-making engine: 23; Intelligent dosing execution unit: 24; Real-time correction unit: 25; Effect Validation and Model Self-Learning Module: 26; Chemical dosing pipeline: 27 Flocculation reaction tank: 28; Formula-Pollution Coefficient Database: 29; Total load forecasting submodule: 30; Material balance model: 31; Component ratio prediction submodule: 32; Formula-component characteristic spectral library: 33. Detailed Implementation

[0031] The detailed features and advantages of this application are described below in the specific embodiments. The content of this description is sufficient to enable any person skilled in the art to understand the technical content of this application and implement it accordingly. Based on the specification, claims and drawings disclosed in this specification, a person skilled in the art can easily understand the related objectives and advantages of this application.

[0032] The invention will now be described with reference to the accompanying drawings, in which similar reference numerals denote similar elements. While specific structures and arrangements are discussed, it should be understood that this is done merely for illustrative purposes. Those skilled in the art will recognize that other structures and arrangements can be used without departing from the spirit and scope of the invention. It will be apparent to those skilled in the art that the invention can also be used in a variety of other applications.

[0033] In this specification and claims, several terms will be used, and unless otherwise indicated, these terms will be defined to have the following meanings: The singular forms “a” and “the” include their corresponding plural forms. “At least one” means one or more, and “more” means two or more. “At least one of the following” or similar expressions refer to any combination of these items, including any combination of single or plural items. For example, at least one of a, b, or c can be expressed as: a, b, c, ab, ac, bc, or abc, where a, b, and c can be single or multiple.

[0034] All figures used to represent component amounts, properties (e.g., molecular weight), reaction conditions, etc., should be considered to be modified in all cases by the terms "within the unavoidable margin of error" or "about". Therefore, the numerical values ​​set forth herein are approximate and may vary depending on the desired properties sought to be obtained by the present invention. The principle of equivalents, which is applied to a minimum and not intended to limit the scope of the claims, should be applied, for example, each value should be interpreted at least according to the specified significant digits and by applying conventional rounding techniques.

[0035] It should be understood that the term "and / or" in this article is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A alone, A and B simultaneously, or B alone. A and B can be singular or plural. Additionally, the character " / " in this article generally indicates an "or" relationship between the preceding and following related objects, but it can also represent an "and / or" relationship. Please refer to the context for a more accurate understanding.

[0036] In the description of this embodiment, it should be noted that the terms "upper", "lower", "inner", "bottom", etc., indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings, or the orientation or positional relationship in which the product is usually placed during use. They are only for the convenience of describing this application and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limitations on this application.

[0037] Unless otherwise indicated, the following abbreviations have the following meanings, and any other abbreviations used herein but not defined have their generally accepted standard meanings: All other terms used herein that are specifically defined herein shall have the general meaning understood by one of ordinary skill in the art, in particular meaning that one of ordinary skill in the art, upon reading the claims, specification and drawings of this patent, can directly and without doubt determine how the technical solution of this patent can be implemented.

[0038] Even if there are incomplete descriptions, omissions, or ambiguities in the grammar, words, punctuation, graphics, symbols, etc. of the claims, specification, and drawings of this patent, a person skilled in the art can still arrive at the only correct understanding by reading the claims, specification, and drawings as a whole without extensive reasoning or experimentation, and effectively exclude various incorrect interpretations that are not aimed at achieving the purpose of this patent.

[0039] Those skilled in the art would first choose to read the claims, specification, and drawings of this patent to reasonably interpret the terms; secondly, they would choose to refer to the relevant definitions in other documents published by the applicant before the filing date to reasonably interpret the terms; thirdly, they would choose the references cited in this patent to reasonably interpret the terms; and finally, they would choose to combine the technical dictionaries, technical manuals, reference books, textbooks, national or industry technical standards, etc., commonly used by those skilled in the art to reasonably interpret the terms.

[0040] To make the objectives, technical solutions, and advantages of this application clearer, the embodiments of this application will be described in further detail below with reference to the accompanying drawings.

[0041] Please refer to Figure 1 This application provides a tobacco wastewater treatment system designed to achieve advanced prediction of tobacco wastewater quality and precise dosing of chemicals. Its core components include a production execution system interface unit 20, an influent spectral fingerprint online monitoring unit 21, a material balance prediction module 22, a dual-mode driven intelligent decision engine 23, an intelligent dosing execution unit 24, a real-time correction unit 25, and an effect verification and model self-learning module 26.

[0042] Each unit and module is connected to a stable data transmission network via industrial Ethernet. The communication protocol adopts the international standard OPCUA protocol, which has cross-platform compatibility, high reliability and data encryption transmission capabilities. It can effectively ensure the real-time, accuracy and stability of data interaction between units / modules, and provide communication guarantee for the efficient operation of the system as a whole.

[0043] The Manufacturing Execution System Interface Unit 20 is deployed inside the industrial control host computer of the wastewater treatment system. It establishes a two-way communication connection with the Manufacturing Execution System (MES system) of the upstream tobacco sheet production line through industrial Ethernet to realize real-time interaction and synchronization of production plan data.

[0044] The core function of this unit is to accurately acquire various production plan data of the upstream sheet production line within a future preset time window, so as to provide a basis for the upstream production conditions for subsequent water quality prediction.

[0045] In this embodiment, the preset time window is preferably set to 2 hours. This duration can be flexibly adjusted according to the actual production scheduling cycle and process adjustment frequency of different tobacco sheet production lines to adapt to the operational needs of different production lines. The acquired production plan data specifically includes key information such as production formula codes (used to associate the pollutant release characteristics corresponding to the formula), production line start and stop signals, planned output corresponding to each production formula, production process switching time points, and production equipment cleaning plans.

[0046] After performing preprocessing operations such as format standardization and abnormal data removal on the acquired raw production plan data, the manufacturing execution system interface unit 20 transmits the standardized data to the material balance prediction module 22 in real time. The data update frequency is synchronized with the plan update frequency of the upstream manufacturing execution system, and the minimum update interval is set to 5 minutes to ensure the timeliness of the production plan data and provide support for the accuracy of water quality prediction.

[0047] The influent spectral fingerprint online monitoring unit 21 is precisely installed on the influent pipe between the main effluent pipe of tobacco sheet production wastewater and the equalization tank of the wastewater treatment system. Its installation position has been strictly calculated, and the hydraulic retention time between it and the subsequent flocculation reaction tank 28 is 30 minutes. This time interval can fully reserve an adjustment and execution window for the pre-dosing of chemicals, ensuring that the dosing of chemicals responds to changes in water quality in a timely manner.

[0048] The core detection equipment of this unit is an online near-infrared spectrometer. In this embodiment, the spectral acquisition range of the online near-infrared spectrometer is set to 900nm to 2500nm, which can fully cover the spectral absorption peaks of characteristic pollutants such as nicotine, lignin, and pectin in tobacco wastewater. Spectral resolution set to 8cm - ¹ It can accurately capture the characteristic spectral differences of different pollutants; the scanning frequency is set to once per minute, with a total of 32 scans per scan and the average value is taken. By averaging multiple scans, the interference of environmental noise, instrument fluctuations and other factors on spectral data can be effectively reduced, thereby improving the stability and accuracy of spectral data.

[0049] After acquiring the raw full-spectral data of tobacco wastewater using an online near-infrared spectrometer, a series of preprocessing operations are required. The preprocessing process consists of Savitzky-Golay smoothing and denoising (eliminating random noise interference), multivariate scattering correction, and baseline correction (eliminating the influence of external factors such as ambient temperature changes, water sample turbidity, and air bubbles in the pipeline on the spectral data). After preprocessing, a standardized full-spectral data matrix is ​​obtained, which can accurately reflect the content characteristics of each component in the wastewater. The preprocessed standardized full-spectral data is synchronously transmitted in real time to the material balance prediction module 22 and the real-time correction unit 25, which are used for auxiliary verification of water quality prediction and dynamic correction of dosage parameters, respectively.

[0050] like Figure 2 As shown, the material balance prediction module 22 is the core calculation unit for water quality advance prediction in the tobacco wastewater treatment system of this application. It integrates two functional sub-units: the total load prediction sub-module 30 and the component ratio prediction sub-module 32. The two sub-modules establish bidirectional data connections with the production execution system interface unit 20 and the influent spectral fingerprint online monitoring unit 21 through the system's internal data bus, respectively, to realize real-time data reception and interaction.

[0051] The total load prediction submodule 30 is primarily responsible for receiving standardized production plan data transmitted from the manufacturing execution system interface unit 20 and calculating and generating predicted influent water quality parameters based on a preset model. The component ratio prediction submodule 32 specifically receives standardized full-spectrum data transmitted from the influent spectral fingerprint online monitoring unit 21 and extracts and outputs real-time influent component feature vectors through feature parsing algorithms. The material balance prediction module 22 integrates the output results of the two submodules, enabling it to predict water quality change trends in advance based on the production plan and correct prediction deviations using real-time spectral data. Ultimately, it accurately outputs the predicted influent water quality parameters, the predicted component feature vectors, and the real-time influent component feature vectors, providing core data support for the subsequent dual-mode driven intelligent decision engine 23 to determine reagent dosing parameters.

[0052] Specifically, the total load prediction submodule 30 is internally configured with a pre-stored formula-pollution factor database 29, a material balance model 31 that has undergone multiple verifications and optimizations, and a feature extraction model. The formula-pollution factor database 29 and the material balance model 31 work together to calculate the predicted influent water quality parameters, ensuring the accuracy and reliability of the prediction results. Among them, the formula-pollution factor database 29, as the basic data support for material balance, needs to be established in advance based on the historical production data and synchronous water quality test data of the corresponding tobacco sheet production line, after scientific fitting analysis, and has dynamic update capabilities, which can be regularly optimized and improved according to the adjustment of production line processes, changes in raw materials, etc.

[0053] The specific process for creating this database is as follows: First, the system collects continuous production operation data of the tobacco sheet production line over the past 12 months, covering key information such as formula code, actual output, production time, raw material input, and process parameters (such as extraction temperature and pulping concentration) for each production batch. At the same time, it collects influent water quality test data of the wastewater generated by the corresponding production batch, focusing on core pollutant indicators such as chemical oxygen demand (COD), nicotine content, and lignin content. The test frequency is synchronized with the production batch to ensure data consistency.

[0054] Subsequently, a linear regression fitting algorithm was used to perform correlation analysis on the collected production data and water quality test data to eliminate the influence of interference factors such as raw material batch differences and environmental temperature fluctuations. Finally, the pollution generation coefficient corresponding to each production formula was calculated. The unit of the pollution generation coefficient was uniformly set as the chemical oxygen demand (COD) generated per ton of product in kilograms (kg / t), which directly reflects the release intensity of pollutants during the production process of different formulas.

[0055] In this embodiment, based on the actual production needs of the corresponding production line, the formula-pollution coefficient database 29 pre-enters the pollution coefficients corresponding to the 8 commonly used production formulas of the production line. Among them, the pollution coefficient corresponding to formula A (60% tobacco stems and 40% broken tobacco leaves) is 185 kg / t, the pollution coefficient corresponding to formula B (40% tobacco stems and 60% broken tobacco leaves) is 210 kg / t, and the pollution coefficient corresponding to formula C (50% tobacco stems, 40% broken tobacco leaves, and 10% tobacco dust) is 160 kg / t. The pollution coefficients of the other 5 commonly used formulas are accurately entered based on the actual linear regression fitting results. The database also supports the functions of manually entering the pollution coefficient of new formulas and automatically updating the original formula coefficients.

[0056] During actual operation, the total load prediction submodule 30 receives standardized production plan data transmitted by the production execution system interface unit 20 in real time, accurately extracts key information such as production formula code, production line start / stop signals, and planned output, calls the corresponding pollution coefficient in the formula-pollution coefficient database 29 according to the formula code, and combines parameters such as planned output and production time to complete the calculation of predicted influent water quality parameters through the material balance model 31. During the calculation process, load fluctuations caused by production line start / stop and process switching are corrected simultaneously, and finally, accurate predicted influent chemical oxygen demand concentration and other core water quality parameters are output, providing data basis for advance control of chemical dosing.

[0057] The mathematical expression for material balance model 31 is:

[0058] in, In the future Predicted influent chemical oxygen demand concentration after a certain time. The set value is 20 minutes, but it can be adjusted within the range of 15 to 30 minutes depending on the hydraulic retention time in the inlet pipe. Indicates the current time Time The yield of this production formula Indicates the first The pollution coefficient corresponding to the production formula In the future Predicted inflow rate after time, This indicates the total number of production formula types. In this embodiment, the influent flow rate is predicted. The wastewater discharge coefficient is obtained by fitting the production line output in the production plan data with the wastewater discharge coefficient corresponding to the historical production data. The wastewater discharge coefficient is the amount of wastewater discharged per ton of product, in cubic meters per ton.

[0059] For example, the current moment At that time, the production line used formula A for production, and the output of formula A was... The corresponding pollution coefficient is 1.2 tons per hour. The predicted inflow rate is 185 kg per ton, and is expected to be within 20 minutes. The predicted influent chemical oxygen demand (COD) concentration for the next 20 minutes is calculated using material balance model 31, with a flow rate of 80 cubic meters per hour. The concentration is 1.2 × 185 / 80 = 2.775 kg per cubic meter, or 2775 mg per liter.

[0060] As one of the core functional sub-units of the material balance prediction module 22, the component ratio prediction sub-module 32 is responsible for accurately predicting and extracting the characteristics of the influent components of tobacco wastewater. This provides core data support at the component level for the subsequent dual-mode driven intelligent decision engine 23 to formulate differentiated reagent dosing strategies. Therefore, this sub-module has a pre-stored formula-component feature spectral library 33. This library is the basic data carrier for realizing advanced prediction of component ratios. Its establishment process has undergone rigorous standard water sample testing, spectral data acquisition, and professional modeling analysis to ensure the accuracy, completeness, and applicability of the spectral library data, which can accurately match the actual production conditions of the corresponding tobacco sheet production line.

[0061] Specifically, the construction process of the formulation-component characteristic spectral library 33 is implemented as follows: The first step is to collect parallel wastewater samples under the corresponding steady-state production conditions for each production formula of the target production line. The number of parallel water samples for each formula should be no less than 50, and the volume of each water sample should be no less than 2L.

[0062] The second step involves conducting laboratory standard tests on the collected water samples to determine the actual content of the four core components: nicotine, lignin, pectin, and protein. The detection methods used were high performance liquid chromatography (HPLC) for nicotine content, ultraviolet spectrophotometry for lignin content, carbazole colorimetric method for pectin content, and Coomassie brilliant blue method for protein content.

[0063] The third step involves using equipment parameters identical to those of an on-site online near-infrared spectrometer to perform a full-spectrum scan on each water sample, thereby acquiring standard full-spectrum data for each sample.

[0064] The fourth step involves using a spectral preprocessing algorithm to preprocess the standard full-spectrum data. The preprocessing steps are Savitzky-Golay smoothing filtering, multivariate scattering correction, standard normal variable transformation, and first-order derivative processing to eliminate irrelevant interference information in the spectral data.

[0065] The fifth step involves using the preprocessed spectral data as the independent variable and the content of the four components obtained from laboratory testing as the dependent variable. A partial least squares discriminant analysis algorithm is used to establish a quantitative correction model between the spectral data and the component content. The model parameters are optimized through cross-validation, and finally, the standard component feature vectors corresponding to each production formula are obtained and stored together in the formula-component feature spectral library 33.

[0066] The predicted component feature vector specifically includes parameters in four core dimensions: nicotine feature value, lignin feature value, pectin feature value, and protein feature value. Each feature value is calculated through the mapping relationship between the actual content of the corresponding component and its spectral characteristics, which can intuitively reflect the relative content and proportion distribution of the corresponding component in the wastewater. Among them, the nicotine feature value can also reflect the feasibility of its recycling. During the actual operation of the system, the component proportion prediction submodule 32 receives standardized production plan data transmitted by the production and manufacturing execution system interface unit 20 in real time through the system's internal data bus, accurately extracts the production formula code, and quickly calls the predicted component feature vector corresponding to the formula code in the formula-component feature spectral library 33. At the same time, it combines the standardized full-spectrum data transmitted by the influent spectral fingerprint online monitoring unit 21 to perform preliminary correction on the predicted component feature vector, eliminate deviations caused by raw material batch differences and minor process fluctuations, and finally outputs an accurate real-time influent component feature vector, completing the advanced prediction of wastewater component proportions and providing reliable component data support for the dynamic adjustment of subsequent reagent dosing strategies.

[0067] The material balance prediction module 22 also includes a feature extraction model (not shown in the figure). This model uses partial least squares discriminant analysis (PLSA) to extract real-time influent component feature vectors from the preprocessed full-spectrum data. The mathematical expression for the feature extraction model using PLS is:

[0068] in, This represents the real-time influent component feature vector, including nicotine feature values, lignin feature values, pectin feature values, and protein feature values. This represents the full-spectrum data matrix after denoising and baseline correction preprocessing. This represents the loading matrix obtained through training with historical spectral data, used to map spectral data to the component feature space. In this embodiment, the loading matrix... The model was trained using 12,000 sets of synchronously collected spectral data from the production line and laboratory component detection data. During the training process, leave-one-out cross-validation was used to optimize the model parameters. The final model's coefficient of determination R² is greater than 0.92, ensuring the accuracy of component feature extraction.

[0069] Please refer to Figure 3 The dual-mode driven intelligent decision-making engine 23, as the core intelligent decision-making unit of the entire tobacco wastewater treatment system, establishes a two-way communication connection with the material balance prediction module 22 through the high-speed data bus inside the system. It can receive the predicted influent water quality parameters (the core of which is the predicted influent COD concentration) and the real-time influent component feature vector (including the feature values ​​of four core components such as nicotine and lignin) output by the module in real time and synchronously. Its core function is to generate dosing control instructions that can be directly issued to the intelligent dosing execution unit 24 based on the above two types of key data through comprehensive analysis, logical judgment and precise calculation. This enables intelligent, differentiated and precise control of the dosing strategy and provides decision support for the efficient operation of the entire wastewater treatment system.

[0070] The dual-mode driven intelligent decision engine 23 integrates two major functional sub-modules: a strategy generation sub-module and a dosage calculation sub-module. The two sub-modules work together seamlessly to ensure the scientific nature and executability of the dosing control instructions. The strategy generation sub-module is responsible for determining "which drug formula to add", while the dosage calculation sub-module is responsible for determining "how much drug to add". Together, they constitute a complete dosing decision system.

[0071] Specifically, the strategy generation submodule internally stores multiple preset processing target modes and a detailed, dynamically updatable drug knowledge base. The two are interconnected and work together to provide basic data and logical basis for the generation of differentiated dosing strategies.

[0072] Among them, the treatment target mode is a core operation mode based on the component characteristics and treatment needs of tobacco wastewater. It includes at least two core modes: nicotine recovery mode and pollutant removal mode. It can be flexibly expanded according to actual production conditions and treatment needs. The original intention of the two modes is to achieve multiple goals of "targeted treatment, resource utilization and cost control".

[0073] The reagent knowledge base is a professional database established after long-term experimental verification and data fitting optimization. It stores basic reagent formulation schemes, reagent addition sequence, addition timing, reagent concentration ratio and key process control parameters that correspond one-to-one with each treatment target mode. It also has dynamic update capability and can regularly optimize reagent formulation parameters based on changes in raw materials, treatment effect feedback and other factors to ensure the suitability and efficiency of reagent addition.

[0074] The basic reagent formulations for the two treatment target modes have been specifically designed to adapt to the characteristics of different influent components, ensuring the maximization of treatment effect and resource recovery benefits.

[0075] The nicotine recovery model prioritizes the recovery of nicotine resources from wastewater while also addressing the removal of other basic pollutants. The corresponding reagent combination primarily uses a calcium-magnesium-aluminum composite flocculant. This type of flocculant exhibits selective adsorption properties, effectively removing suspended impurities and some colloidal pollutants from wastewater while minimizing the adsorption and removal of nicotine components, facilitating subsequent nicotine separation and recovery. Simultaneously, an appropriate amount of micro-sand suspension is added as a coagulant aid. The micro-sand acts as a flocculation core, accelerating floc formation and sedimentation, shortening the flocculation reaction time, improving flocculation efficiency, reducing the dosage of the main reagent, and lowering treatment costs.

[0076] The pollutant removal mode focuses on "efficiently removing pollutants such as COD, lignin, and pectin from wastewater to ensure that the effluent meets the standards." The corresponding reagent combination uses polyaluminum chloride (PAC) as the core flocculant. This reagent has advantages such as strong adsorption and bridging ability, wide applicability to water quality range, and fast floc formation, and can efficiently remove recalcitrant organic pollutants and colloidal substances from wastewater.

[0077] Simultaneously, anionic polyacrylamide (APAM) is used as a coagulant aid to enhance the strength and settling velocity of the flocs, further improving the pollutant removal efficiency, ensuring the treatment effect of the physicochemical pretreatment stage, and laying the foundation for the stable operation of the subsequent biochemical system.

[0078] It is understandable that the pattern matching logic of the strategy generation submodule adopts a precise threshold comparison method to ensure the scientific and timely nature of pattern switching. The specific matching process is as follows: First, the strategy generation submodule receives the real-time influent component feature vector transmitted by the material balance prediction module 22, accurately extracts the nicotine feature value from it, and converts the nicotine feature value into the corresponding actual nicotine concentration (unit: mg / L) through a preset mapping algorithm.

[0079] Subsequently, the calculated actual nicotine concentration is precisely compared with the system's preset nicotine concentration threshold. This preset threshold has been determined through extensive testing and is preferably set to 200 mg / L in this embodiment. This threshold can be flexibly adjusted according to the nicotine production, recovery value, and treatment standards of different production lines. When the comparison result shows that the actual nicotine concentration is ≥200 mg / L, the system determines that the current influent nicotine content is high and has significant resource recovery value. It automatically matches and activates the nicotine recovery mode, and simultaneously quickly retrieves the corresponding basic reagent formulation, dosing sequence, and related process parameters from the reagent knowledge base. For example, the basic reagent formulation for the nicotine recovery mode uses calcium magnesium aluminum composite flocculant as the main reagent, supplemented by micro-sand suspension as a coagulant aid.

[0080] When the comparison result shows that the actual nicotine concentration is <200mg / L, the system determines that the current influent nicotine content is too low and has no economic recovery value. Prioritizing pollutant removal efficiency, the system automatically matches and activates the pollutant removal mode, simultaneously retrieving the corresponding basic reagent formulation. This allows for dynamic mode switching based on differences in influent components, balancing environmental treatment and resource recovery. For example, the basic reagent formulation for the pollutant removal mode uses polyaluminum chloride as the main flocculant and anionic polyacrylamide as the coagulant aid.

[0081] The dosage calculation submodule incorporates a predictive dose response model. This model, a core tool for accurate drug dosage calculation, is scientifically established in advance through orthogonal experiments combined with multivariate nonlinear regression fitting to ensure the accuracy and applicability of the model's calculation results, enabling it to adapt to drug dosage requirements under different operating conditions. The specific establishment process is as follows: For the two pre-set treatment target modes (nicotine recovery mode and pollutant removal mode) in the strategy generation submodule, a multivariate orthogonal flocculation experiment was designed with corresponding basic reagent formulation schemes. The experimental variables mainly cover the two core influencing factors of influent flow rate and influent chemical oxygen demand (COD) concentration, while also taking into account secondary interference factors such as water temperature and pH value. The optimal reagent dosing rate data under different operating conditions were obtained through system experiments.

[0082] Subsequently, a multivariate nonlinear regression fitting algorithm was used to fit and analyze the correlation data between the influent flow rate, influent COD concentration and optimal reagent dosing acceleration rate obtained from the experiment. This eliminated the influence of experimental errors and interference factors, established an accurate mapping relationship among the three, and finally formed a predictive dose response model that can be directly used for actual operation calculations.

[0083] During actual system operation, the dosage calculation submodule receives the predicted influent water quality parameters output by the material balance prediction module 22 in real time, accurately extracts the predicted influent flow rate and predicted influent COD concentration, and simultaneously obtains the basic reagent formulation scheme retrieved by the strategy generation submodule. Based on the reagent formulation type, the matching model corresponds to the calculation parameters, and the predictive dosage response model is called for accurate calculation. Finally, a scientific and reasonable predicted reagent dosing rate is output, providing a precise dosage control basis for the intelligent dosing execution unit 24.

[0084] In this embodiment, under the pollutant removal mode, the mathematical expression of the predictive dose-response model for polyaluminum chloride is:

[0085] in, The predicted dosing rate of polyaluminum chloride is expressed in L / hour. To predict the inflow rate, the unit is cubic meters per hour. To predict the influent chemical oxygen demand concentration, the unit is mg per liter.

[0086] The mathematical expression for the predicted dosing rate of polyacrylamide is:

[0087] in, The predicted dosing rate of polyacrylamide is expressed in L / hour. The inflow rate is predicted in cubic meters per hour.

[0088] Under the nicotine recovery model, the mathematical expression for the predictive dose-response model of the calcium-magnesium-aluminum composite flocculant is as follows:

[0089] in, The predicted dosing rate of calcium-magnesium-aluminum composite flocculant is expressed in L / hour. To predict the inflow rate, the unit is cubic meters per hour. The value represents the nicotine concentration corresponding to the real-time influent component feature vector, in mg per liter.

[0090] The mathematical expression for the predicted dosing acceleration rate of microsand suspension is:

[0091] in, The predicted dosing rate of the microsand suspension is expressed in L / hour. The inflow rate is predicted in cubic meters per hour.

[0092] The dosing control instructions generated by the dual-mode driven intelligent decision engine 23 include the predicted dosing rate and the corresponding dosing execution time. The dosing execution time is determined based on the hydraulic retention time of the wastewater from the spectral monitoring point to the flocculation reaction tank 28. In this embodiment, the dosing execution time is set to 5 minutes before the wastewater actually arrives at the flocculation reaction tank 28 to ensure the timing accuracy of the pre-dosing operation.

[0093] Specifically, the real-time correction unit 25 is connected to the influent spectral fingerprint online monitoring unit 21 and the dual-mode driven intelligent decision engine 23. Its core function is to calculate the deviation between the real-time influent component feature vector and the predicted component feature vector, generate correction coefficients based on the deviation, and adjust the generated dosing control commands in real time. The real-time correction unit 25 first calculates the deviation between the real-time influent component feature vector and the predicted component feature vector. The deviation is calculated using the Euclidean distance between the two vectors, and the mathematical expression is:

[0094] in, For the current moment The deviation value, to These are the nicotine, lignin, pectin, and protein characteristic values ​​in the real-time influent component feature vector, respectively. to These are the feature values ​​of the four dimensions corresponding to the feature vector of the predicted component.

[0095] The real-time correction unit 25 incorporates an incremental proportional-integral-derivative (PID) control algorithm. Based on the calculated deviation value, it generates a correction amount for the drug dosing acceleration rate. The mathematical expression of the incremental PID control algorithm is:

[0096] in, Indicates the current time Correction amount for drug dosing rate. Indicates the current time The deviation value, This represents the deviation value at the previous moment. This represents the deviation value between the two previous moments. Represents the proportionality coefficient. Represents the integral coefficient. This represents the differential coefficient. In this embodiment, the proportional coefficient... The value is 0.65, and the integral coefficient is... The value is 0.12, and the differential coefficient is... The value is set to 0.08, and the upper and lower limits of the correction amount are set to ±20% of the predicted dosing rate to avoid drastic fluctuations in the dosing rate and ensure the stability of the dosing process.

[0097] The dual-mode driven intelligent decision engine 23 adjusts the generated dosing control command in real time according to the correction amount generated by the real-time correction unit 25. The adjustment cycle is consistent with the acquisition cycle of spectral data, which is once per minute, to achieve closed-loop control of predictive execution and precise correction.

[0098] Specifically, the intelligent dosing execution unit 24 is connected to the dual-mode driven intelligent decision engine 23, and its core function is to adjust the type and rate of dosing of various agents according to the dosing control commands. The intelligent dosing execution unit 24 includes four independent agent dosing pipelines 27, each equipped with a variable frequency metering pump, a chemical storage tank, a pulsation damper, and a flow transmitter. The four sets of agent dosing pipelines 27 correspond to the dosing of calcium magnesium aluminum composite flocculant, polyaluminum chloride, polyacrylamide, and microsand suspension, respectively.

[0099] In this embodiment, the flow rate adjustment range of the variable frequency metering pump is from 0L / hour to 500L / hour, with an adjustment accuracy of ±0.5%. The measurement accuracy of the flow transmitter is ±0.2%. It can collect the actual dosing flow rate of the agent in real time and send the collected data back to the dual-mode driven intelligent decision engine 23 to form the underlying closed-loop control of the agent dosing flow rate.

[0100] After receiving the dosing control command, the intelligent dosing execution unit 24 initiates a pre-dosing operation to predict the acceleration rate of the dosing agent at a first preset time before the wastewater actually reaches the flocculation reaction tank 28. In this embodiment, the first preset time is set to 5 minutes to ensure that the agent has already completed preliminary mixing with the water when the wastewater enters the flocculation reaction tank 28, achieving differentiated dosing synchronized with changes in water quality. Upon receiving the correction command generated by the real-time correction unit 25, the intelligent dosing execution unit 24 adjusts the corresponding agent dosing acceleration rate in real time by regulating the operating frequency of the variable frequency metering pump, thus completing the real-time correction of the dosing command.

[0101] Specifically, the effect verification and model self-learning module 26 is connected to the material balance prediction module 22 and the dual-mode driven intelligent decision engine 23, respectively. Its core function is to collect data from the entire dosing process and periodically update the pollution generation coefficient database and the reagent knowledge base to continuously optimize the system's control performance. The effect verification and model self-learning module 26 collects production plan data, full-spectrum data, dosing control commands, actual dosing flow rate data, influent water quality online monitoring data, and final effluent water quality online monitoring data for each dosing process. The time resolution of all data acquisition is 1 minute, and the data is stored in the system's industrial control database.

[0102] In this embodiment, the effect verification and model self-learning module 26 updates the pollution generation coefficient database and the reagent knowledge base periodically, with a 24-hour update cycle. For the pollution generation coefficient database, within each update cycle, the continuous production data collected during that cycle and the corresponding actual influent water quality detection data are used to refit the pollution generation coefficients corresponding to each production formula through linear regression. When the deviation between the newly fitted pollution generation coefficient and the corresponding coefficient in the original database exceeds 10%, the new pollution generation coefficient replaces the data in the original database, improving the accuracy of total load prediction. For the reagent knowledge base, within each update cycle, the dosing control data collected during that cycle and the corresponding effluent water quality data are used through orthogonal experimental optimization analysis to update the ratio parameters of the basic reagent formulation scheme and the coefficients of the predictive dose response model, ensuring continuous optimization of the control effect of reagent dosing.

[0103] Please refer to Figure 4 This application clarifies the control logic switching nodes and parameter adjustment rules for each stage, from the triggering of non-steady-state signals in the production plan to the setting of full-time control nodes and time windows for prediction and control, real-time correction, addition execution, and effect verification.

[0104] Specifically, in addition to acquiring regular production plan data, the Manufacturing Execution System (MES) interface unit 20 also sets up dedicated signal acquisition and priority processing logic for non-steady-state operating conditions. The MES interface unit 20 collects three types of non-steady-state characteristic signals in real time: formula change warning signals from upstream production lines, production line start / stop trigger signals, and equipment cleaning reservation signals. These signals have a higher acquisition priority than regular production plan data, and the acquisition update interval is set to 1 minute to ensure that trigger signals for non-steady-state operating conditions are captured by the system first. It is understandable that the fluctuation amplitude and rate of change of wastewater quality under non-steady-state operating conditions are significantly higher than under steady-state production conditions; therefore, it is necessary to shorten the data update interval and improve the system's response speed.

[0105] For example, when the manufacturing execution system interface unit 20 receives a formula switching warning signal from the upstream production line, it immediately extracts the two production formula codes before and after the switching, the switching execution time point, and the output change curve data during the switching process, and transmits them synchronously to the material balance prediction module 22 to provide complete pre-data support for water quality prediction under non-steady-state conditions.

[0106] Specifically, the material balance prediction module 22 sets up a dynamic sliding time window prediction mechanism for unsteady-state operating conditions, replacing the fixed time window prediction under steady-state operating conditions. Upon receiving an unsteady-state characteristic signal, the total load prediction submodule 30 adjusts the prediction time window... The steady-state time window has been adjusted from 20 minutes to 15 minutes, and the prediction calculation time step has been shortened from 5 minutes to 1 minute, enabling continuous minute-by-minute prediction of water quality changes during unsteady-state processes. It is understandable that shortening the prediction time window and calculation step can effectively reduce prediction errors caused by yield fluctuations during unsteady-state processes, improving the timeliness and accuracy of prediction results.

[0107] Under non-steady-state conditions, the control logic of the dual-mode driven intelligent decision engine 23 is divided into three stages: pre-control decision stage, transition adjustment stage, and steady-state maintenance stage. The control parameters and decision logic of the three stages are switched sequentially to ensure a smooth transition of dosing control during the switching of operating conditions.

[0108] Specifically, the pre-control decision-making stage lasts from 15 minutes before the formula switching execution time to the switching execution time. During this stage, the dual-mode driven intelligent decision engine 23 generates the dosing control instructions for the transition stage in advance based on the predicted influent water quality parameters and predicted component feature vectors corresponding to the two production formulas before and after the switch. The intelligent dosing execution unit 24 initiates the pre-dosing operation 5 minutes before the switching execution time to complete the transition adjustment of the dosing strategy in advance. The transition adjustment stage lasts from the formula switching execution time to 30 minutes after the switch is completed. During this stage, the strategy generation submodule increases the update frequency of the real-time influent component feature vector from once per minute to once every 30 seconds, the dosage calculation submodule simultaneously shortens the prediction calculation step size, and the real-time correction unit 25 adjusts the proportional coefficient of the incremental proportional integral derivative control algorithm. The integral coefficient was adjusted from 0.65 to 0.8. The differential coefficient was adjusted from 0.12 to 0.08. The value was adjusted from 0.08 to 0.1 to improve the response speed to rapid fluctuations. Simultaneously, the upper and lower limits of the correction amount were set to ±15% of the predicted dosing acceleration rate to avoid significant jumps in the dosing rate during operating condition switching. The steady-state maintenance phase occurs 30 minutes after the formulation switch is completed. During this phase, the system automatically reverts to the control parameters corresponding to the steady-state production condition and executes the conventional predictive and corrective control logic.

[0109] For example, when the production line switches from formula A to formula B, the treatment target mode corresponding to formula A is the pollutant removal mode, and the treatment target mode corresponding to formula B is the nicotine recovery mode. After receiving the formula switching warning signal, the dual-mode driven intelligent decision engine 23 completes the preloading of the reagent formula schemes corresponding to the two modes 15 minutes in advance, starts the pre-addition of calcium magnesium aluminum composite flocculant 5 minutes before the switching execution time, and gradually reduces the dosing rate of polyaluminum chloride and polyacrylamide to achieve a smooth transition between the two dosing strategies and avoid problems such as reagent dosing gaps or overdosing during the switching process.

[0110] like Figure 5As shown, the complete iterative process from full-process data collection, data preprocessing, model validation, parameter optimization to database updates is clearly demonstrated, and the execution standards and judgment rules for each step are clearly defined.

[0111] Specifically, the iterative update process of the effect verification and model self-learning module 26 is implemented as follows.

[0112] The first step is full-process data acquisition. The module collects full-process data under all operating conditions within a 24-hour iteration cycle, including production plan data, non-steady-state characteristic signal data, full-spectrum data, predicted influent water quality parameters, real-time influent component feature vectors, dosing control commands, actual chemical dosing flow data, influent water quality online monitoring data, effluent water quality online monitoring data, and laboratory manual testing data. All data are aligned using a unified timestamp with a time resolution of 1 minute.

[0113] The second step is data preprocessing, which involves outlier removal and data standardization of the collected data. The Laida criterion is used to remove outliers from the data, with the removal criterion being that the deviation of the data from the mean exceeds 3 times the standard deviation. Linear interpolation is then used to complete the data after outlier removal to ensure the integrity and validity of the dataset.

[0114] The third step is to verify the model performance. The performance of the total quantity prediction model and component feature extraction model of the material balance prediction module 22 and the dose response model of the dual-mode driven intelligent decision engine 23 are verified respectively. The average relative error between the predicted value and the actual detection value of each model is calculated.

[0115] The fourth step involves parameter optimization and database updates. When the average relative error of the total quantity prediction model exceeds 10%, the pollution generation coefficient and non-steady-state operating condition correction factor of the corresponding production formula are refitted, and the formula-pollution generation coefficient database is updated. When the average relative error of the component feature extraction model exceeds 15%, the partial least squares discriminant analysis model is incrementally trained using newly added effective samples, and the loading matrix is ​​updated. The formulation-component characteristic spectrum library 33 is used. When the effluent water quality compliance rate of the dose response model is lower than the preset threshold, the coefficients of the dose response model are refitted, and the basic drug formulation schemes and dosage calculation parameters in the drug knowledge base are updated.

[0116] The fifth step is model iteration archiving. The updated model parameters and database content are archived and stored, while the model parameters and database content of the previous version are retained to ensure that the system can be rolled back and to avoid system failures caused by abnormal updates.

[0117] Specifically, in this embodiment, the preset threshold for the effluent quality compliance rate is set to 95%. The compliance criteria are that the chemical oxygen demand (COD) removal rate of the effluent from the flocculation pretreatment unit is not less than 40%, the suspended solids removal rate is not less than 50%, and the nicotine recovery rate under the nicotine recovery mode is not less than 55%. It is understandable that setting a reasonable compliance rate threshold ensures that the model's self-learning process always revolves around the system's core processing objectives, avoiding meaningless parameter iterations.

[0118] In some optional implementations, the system also incorporates abnormal operating condition protection logic to ensure stable operation even in the event of monitoring equipment failure, communication interruption, or extreme water quality anomalies. Specifically, when the influent spectral fingerprint online monitoring unit 21 experiences equipment failure or data anomalies, the system automatically switches to a pure feedforward predictive control mode, generating dosing control commands solely based on production plan data via the material balance prediction module 22, while simultaneously triggering an equipment failure alarm to alert maintenance personnel for repairs. When the manufacturing execution system interface unit 20 experiences a communication interruption and is unable to obtain upstream production plan data, the system automatically switches to a pure feedback control mode, generating dosing control commands based on real-time full-spectrum data collected by the influent spectral fingerprint online monitoring unit 21, and simultaneously performing closed-loop correction based on effluent water quality data via the real-time correction unit 25 to ensure stable treatment results. When the influent chemical oxygen demand (COD) concentration exceeds the preset upper limit threshold of 15,000 mg / L, the system automatically triggers high-load protection logic, increasing the dosing rate by a preset ratio based on the basic reagent formulation, while extending the flocculation reaction time to ensure treatment effectiveness under high-load conditions.

[0119] This embodiment conducts continuous operation verification under three typical non-steady-state conditions in a tobacco sheet production line: formula switching, production line start-up and shutdown, and equipment cleaning. The verification results show that the system's average relative error in predicting changes in influent water quality under non-steady-state conditions does not exceed 10%, there is no significant control lag during the switching process of the chemical dosing strategy, and the effluent water quality compliance rate of the flocculation pretreatment unit remains stable at over 96%. Compared with the traditional fixed-value dosing method, the average amount of chemical dosing under non-steady-state conditions is reduced by 15%, effectively solving the problems of control lag, unstable treatment effect, and chemical waste in existing technologies under sudden changes in water quality.

[0120] Please refer to Figure 6 Based on the same inventive concept, this application also provides a method for treating tobacco wastewater, which can be applied to the above-mentioned tobacco wastewater treatment system. The specific steps of the method are as follows: Step S1: Obtain standard tobacco production plan data and standard tobacco wastewater full-spectrum data.

[0121] Specifically, step S1 includes: Step A1: Obtain tobacco production plan data within the preset time window.

[0122] Through the Manufacturing Execution System (MES) interface unit 20, the raw data of the tobacco production plan within a preset time window is obtained from the upstream tobacco sheet production line's Manufacturing Execution System (MES system). In this embodiment, the preset time window is preferably 2 hours. This duration can be flexibly adapted according to the actual scheduling cycle and process adjustment frequency of the production line to ensure that the obtained production plan data can provide effective support for subsequent water quality advance prediction.

[0123] Step A2: Perform the first preprocessing on the tobacco production plan data to obtain standard tobacco production plan data.

[0124] The acquired raw tobacco production plan data undergoes a first preprocessing operation to ultimately obtain standardized tobacco production plan data (i.e., standard tobacco production plan data). This first preprocessing operation includes at least one or more of the following: normalization, dimensionless formatting, outlier removal, and data format standardization. Its core purpose is to eliminate interfering factors in the raw data, standardize data specifications, and ensure that the subsequent material balance prediction module 22 can accurately analyze and efficiently utilize the data, thereby guaranteeing the accuracy of water quality prediction results.

[0125] Step A3: Obtain full-spectral data of tobacco wastewater that has not yet flowed into the wastewater treatment process.

[0126] The influent spectral fingerprint online monitoring unit 21 collects raw full-spectrum data of tobacco wastewater that has not yet flowed into the main treatment unit of the wastewater treatment system.

[0127] Step A4: Perform a second preprocessing on the full-spectrum data of tobacco wastewater to obtain standard full-spectrum data of tobacco wastewater.

[0128] A second preprocessing operation is performed on the collected raw full-spectral data of tobacco wastewater to obtain the final full-spectral data of standard tobacco wastewater. This second preprocessing operation includes at least one of smoothing and denoising or baseline correction, which can be flexibly selected based on the interference situation of the actual spectral data. The core purpose is to eliminate noise, baseline drift, and other interference factors in the raw spectral data, thereby improving the accuracy and reliability of the spectral data.

[0129] Specifically, the influent spectral fingerprint online monitoring unit 21 collects the raw full-spectrum data of the tobacco wastewater that has not yet flowed into the main treatment unit of the wastewater treatment system in real time. The collected raw full-spectrum data is then subjected to a second preprocessing operation, such as smoothing and denoising, and baseline correction. After processing, standard tobacco wastewater full-spectrum data is obtained. This standard tobacco wastewater full-spectrum data can accurately reflect the content characteristics of each core component in the wastewater. Subsequently, it is transmitted in real time to the material balance prediction module 22 and the real-time correction unit 25, respectively, for auxiliary verification of water quality prediction and dynamic correction of dosing parameters.

[0130] Step S2: Process the tobacco production plan data according to the material balance model 31 to obtain the predicted influent water quality parameters, including the predicted influent chemical oxygen demand concentration.

[0131] Specifically, the standard tobacco production plan data includes the output of the production formula, the pollution coefficient of the production formula, and the influent flow rate. The calculation formula of material balance model 31 includes:

[0132] in, In the future Predicted influent chemical oxygen demand concentration after a certain time. Indicates the current time Time The yield of this production formula Indicates the first The pollution coefficient corresponding to the production formula In the future Predicted inflow rate after time, This indicates the total number of production formulas.

[0133] It is understandable that, based on the material balance model 31, the standard tobacco production plan data obtained in step S1 is systematically analyzed and accurately calculated to finally obtain the predicted influent water quality parameters. Among them, the core indicator of the predicted influent water quality parameters is the predicted influent chemical oxygen demand (COD) concentration. At the same time, the predicted concentrations of core pollutants such as nicotine and lignin can be output simultaneously according to the actual treatment needs, providing comprehensive water quality load data support for subsequent reagent dosing decisions.

[0134] First, the total load prediction submodule 30 inside the material balance prediction module 22 receives the standard tobacco production plan data after preprocessing in step S1 in real time, and accurately extracts key core parameters such as production formula code, planned output, and production time.

[0135] Subsequently, based on the extracted production formula code, the corresponding pollution coefficient in the built-in formula-pollution coefficient database 29 is quickly retrieved. This pollution coefficient is determined by linear regression fitting optimization based on the historical production data of the corresponding production line and the synchronous water quality test data, and can accurately reflect the release intensity of pollutants during the production process of the formula.

[0136] Next, the retrieved pollution generation coefficient, planned output, production time and other parameters are substituted into the preset material balance model 31. At the same time, the load fluctuations caused by factors such as production line start-up and shutdown and process switching are automatically corrected to complete the accurate calculation of the predicted influent water quality parameters.

[0137] Finally, the predicted influent water quality parameters for the next 20 minutes are output. This time point is determined through rigorous hydraulic calculations and matches the installation location of the influent spectral fingerprint online monitoring unit 21 and the timing of the reagent dosing. This ensures the foresight of the predicted data and allows sufficient time for subsequent adjustments to the reagent dosing, achieving precise linkage between advanced water quality prediction and reagent dosing, and ensuring the stability of the pretreatment effect.

[0138] Step S3: Process the full-spectrum data of tobacco wastewater according to the feature extraction model to obtain the real-time influent component feature vector.

[0139] Specifically, the real-time influent component feature vector includes nicotine feature values, lignin feature values, pectin feature values, and protein feature values. The calculation formula for the feature extraction model includes:

[0140] in, This represents the feature vector of real-time influent components. This represents the full spectral data of standard tobacco wastewater. This represents the load matrix obtained through training with historical spectral data.

[0141] Based on the feature extraction model, a systematic feature analysis and precise processing of the full-spectral data of standard tobacco wastewater is performed to obtain a real-time influent component feature vector that accurately reflects the composition of the wastewater. This vector serves as the core component basis for the subsequent dual-mode driven intelligent decision engine 23 to switch processing modes and formulate differentiated reagent dosing strategies, directly determining the targeting and rationality of reagent dosing. It can be understood that this application combines the dual approach of "formula prediction + spectral measurement" to achieve accurate identification and quantification of wastewater components through the feature extraction model, overcoming the limitations of single prediction or single detection, and ensuring the accuracy and reliability of the component feature vector. The specific implementation process is as follows: First, the component ratio prediction submodule 32 receives standard tobacco production plan data, accurately extracts the production formula code, and quickly retrieves the corresponding predicted component feature vector from its internally stored formula-component feature library 33 based on the formula code. This predicted component feature vector is pre-modeled based on the standard operating conditions of the production formula and includes four core dimension parameters: nicotine feature value, lignin feature value, pectin feature value, and protein feature value. It can initially reflect the component ratio benchmark of the production wastewater of the formula, providing a reference benchmark for subsequent real-time component extraction.

[0142] Secondly, the full-spectral data of standard tobacco wastewater obtained after the second preprocessing operation is input into a preset feature extraction model. This feature extraction model is constructed using the partial least squares discriminant analysis (PLS-DA) algorithm, which has extremely strong feature extraction and noise reduction capabilities. It can accurately capture the characteristic spectral signals corresponding to each component in complex spectral data and effectively avoid interference caused by the overlap of different component spectra.

[0143] This algorithm performs feature analysis on the standardized full-spectrum data matrix, extracting the characteristic peak signals corresponding to the four core components: nicotine, lignin, pectin, and protein. Combined with the preset spectrum-component mapping relationship, the characteristic peak signals are converted into corresponding component feature values, thus initially generating component feature vectors based on real-time spectral detection.

[0144] The feature values ​​of each component in this vector are positively correlated with the actual content of the corresponding component in the wastewater, which can intuitively reflect the relative content and proportion distribution of each component. Among them, the nicotine feature value can be directly used as the core judgment basis for subsequent pattern matching, while the lignin, pectin and protein feature values ​​are used to help optimize the dosage of the agent and ensure that the agent dosage is accurately matched with the wastewater components.

[0145] Step S4: Determine the reagent dosing parameters for tobacco wastewater based on the predicted influent water quality parameters and the real-time influent component feature vector.

[0146] The drug delivery parameters include the type of drug and the drug delivery rate. Step S4 includes: Step S41: Match the real-time influent component feature vector with the preset treatment target mode to determine the type of reagent to be added.

[0147] Step S42: Determine the dosing rate of the chemical based on the predicted influent flow rate, predicted influent water quality parameters, and the type of chemical to be added.

[0148] First, the real-time influent component feature vectors are compared and matched one by one with the preset treatment target mode. Based on the matching results, the corresponding basic reagent formula scheme is quickly retrieved from the reagent knowledge base to clarify the reagent type, dosing order and basic concentration ratio.

[0149] Secondly, the predicted influent flow rate, predicted influent water quality parameters (primarily the predicted influent COD concentration), and the retrieved basic reagent formulation are simultaneously input into the dosage calculation submodule. This submodule then calls a preset predictive dosage response model to perform precise calculations and obtain the optimal predicted reagent dosing rate.

[0150] Finally, combining the installation location of the influent spectral fingerprint online monitoring unit 21 with the hydraulic retention time of the flocculation reaction tank 28, the pre-dosing operation of the reagent is started 5 minutes before the wastewater actually reaches the flocculation reaction tank 28. By adjusting the working frequency of the corresponding strain frequency metering pump, the type and rate of reagent addition are precisely controlled to ensure that the reagent and wastewater are fully mixed and reacted, thus ensuring the physical and chemical pretreatment effect and achieving advanced pre-control and precise matching of reagent addition.

[0151] To further improve the wastewater treatment effect, steps S43 to S45 can be performed, as follows: Step S43: Based on the production formula code of the standard tobacco production plan data, call the pre-stored formula-component feature spectrum library 33 to generate the predicted component feature vector.

[0152] Step S44: Calculate the deviation data between the real-time influent component feature vector and the predicted component feature vector, and obtain the deviation correction coefficient based on the deviation data.

[0153] Step S45: Input the deviation correction coefficient into the incremental proportional-integral-derivative control algorithm to adjust the type of drug and / or the drug dosing rate.

[0154] The incremental proportional-integral-derivative control algorithm is as follows:

[0155] in, Indicates the current time Correction amount for drug dosing rate. Indicates the current time The deviation between the real-time influent component feature vector and the predicted component feature vector. This represents the deviation value at the previous moment. This represents the deviation value between the two previous moments. Represents the proportionality coefficient. Represents the integral coefficient. This represents the differential coefficient.

[0156] It is understandable that by generating a correction amount for the drug dosing acceleration rate through an incremental proportional-integral-derivative control algorithm, the dosing control command being executed is adjusted in real time, thereby adjusting the corresponding drug dosing acceleration rate and completing the closed-loop correction of the dosing control.

[0157] The actual operational results of this embodiment are as follows: Under normal production and formula switching conditions of the papermaking tobacco sheet production line, this method can predict changes in influent water quality up to 20 minutes in advance, with an average relative error of less than 8% in predicting the influent chemical oxygen demand concentration and an accuracy rate of more than 90% in identifying influent component characteristics. When formula switching in the production line causes sudden changes in water quality, this method can complete the switching and pre-dosing of chemical dosing strategies in advance, avoiding the fluctuations in treatment effect caused by traditional delayed control.

[0158] Compared with the traditional fixed-value dosing method, the average dosage of chemicals in this system is reduced by 18%, the chemical oxygen demand removal rate of the effluent from the flocculation pretreatment unit is stable at over 45%, and the nicotine recovery rate under the nicotine recovery mode reaches over 60%, effectively reducing the operating cost of wastewater treatment, while realizing the recovery and utilization of valuable components in wastewater.

[0159] In this specification, references to "an embodiment" or "a specific implementation" mean that a particular feature, structure, or characteristic described in connection with that embodiment / specific implementation is included in at least one embodiment / specific implementation of the invention. Therefore, the phrase "in one embodiment / specific implementation" appearing in various places in this specification does not necessarily refer to the same embodiment / setting, but rather to potentially different embodiments. Furthermore, specific features, structures, or characteristics may be combined in one or more embodiments / settings in any suitable manner, as will be apparent to those skilled in the art from this disclosure.

[0160] Similarly, it should be understood that in the above description of exemplary embodiments / specific implementations of the invention, various features of the invention are sometimes combined in a single embodiment / specific implementation or its figures and description, with the aim of simplifying the disclosure and aiding in the understanding of one or more of the various aspects of the invention. However, the method of description in this patent should not be construed as reflecting an intention that the claimed features of the invention are more than those expressly stated in each claim, except where explicitly stated otherwise or in obvious technical contradiction or exclusion. Rather, the inventive aspect reflected in the claims lies in not all the features of a single foregoing disclosed embodiment / specific implementation. Therefore, the claims following the detailed description are expressly incorporated herein by reference, each claim existing independently as a separate embodiment / specific implementation of the invention.

[0161] Furthermore, while some embodiments / specific implementations described herein include, but are not limited to, other features included in other embodiments / specific implementations, combinations of features from different embodiments / specific implementations are intended to be within the scope of the invention and form different embodiments / specific implementations, as will be understood by those skilled in the art. For example, in the following claims, embodiments / specific implementations of any claim can be used in any combination.

[0162] The terms and expressions used in this specification are for illustrative purposes and not for limitation. In using these terms and expressions, it is not intended to exclude any equivalents of the features or portions thereof shown and described, but rather to recognize that various modifications may be possible within the scope of the invention.

[0163] Therefore, it should be understood that although the invention has been specifically disclosed through preferred embodiments, exemplary embodiments and optional features, those skilled in the art may take variations or modifications of the concepts disclosed herein, and such variations and modifications are therefore considered to be within the scope of the invention as defined by the appended claims.

[0164] The specific embodiments given in this specification are examples of useful implementations of the present invention. It will be apparent to those skilled in the art that the present invention can be implemented using many variations of the devices, device components, and method steps disclosed in this specification.

[0165] The foregoing description of specific embodiments fully discloses the general features of the present invention, enabling others to easily modify and / or adapt such specific embodiments for various applications by applying knowledge within the scope of the art, without conducting excessive experimentation and without departing from the general concept of the present invention.

[0166] Therefore, based on the teachings and guidance provided herein, it is intended that such modifications and alterations be included within the meaning and scope of equivalents of the disclosed embodiments. It should be understood that the wording or terminology used herein is for descriptive purposes and is not intended to be limiting; thus, the wording or terminology in this specification will be interpreted by those skilled in the art based on the foregoing teachings and guidance.

[0167] Furthermore, the scope of the invention should not be limited to any of the exemplary embodiments described above, but only to the appended claims and their equivalents.

Claims

1. A method for treating tobacco wastewater, characterized in that, The tobacco wastewater treatment method includes: Step S1: Obtain standard tobacco production plan data and standard tobacco wastewater full-spectrum data; Step S2: Process the tobacco production plan data according to the material balance model to obtain the predicted influent water quality parameters, including the predicted influent chemical oxygen demand concentration. Step S3: Process the full-spectrum data of the tobacco wastewater according to the feature extraction model to obtain the real-time influent component feature vector; Step S4: Determine the reagent dosing parameters for tobacco wastewater based on the predicted influent water quality parameters and the real-time influent component feature vector.

2. The method for treating tobacco wastewater according to claim 1, characterized in that, Step S1 includes: Step A1: Obtain tobacco production plan data within the preset time window; Step A2: Perform a first preprocessing on the tobacco production plan data to obtain the standard tobacco production plan data. The first preprocessing includes at least one of the following: normalization, dimensionless processing, outlier removal, and data format unification.

3. The method for treating tobacco wastewater according to claim 1, characterized in that, Step S1 includes: Step A3: Obtain full-spectral data of tobacco wastewater that has not yet flowed into the wastewater treatment process; Step A4: Perform a second preprocessing on the full-spectrum data of the tobacco wastewater to obtain the full-spectrum data of the standard tobacco wastewater. The second preprocessing includes at least one of smoothing and noise reduction and baseline correction.

4. The method for treating tobacco wastewater according to claim 1, characterized in that, The standard tobacco production plan data includes the production output of the production formula, the pollution coefficient of the production formula, and the influent flow rate. The calculation formula of the material balance model includes: in, In the future Predicted influent chemical oxygen demand concentration after a certain time. Indicates the current time Time The yield of this production formula Indicates the first The pollution coefficient corresponding to the production formula In the future Predicted inflow rate after time, This indicates the total number of production formulas.

5. The method for treating tobacco wastewater according to claim 1, characterized in that, The real-time influent component feature vector includes nicotine feature values, lignin feature values, pectin feature values, and protein feature values. The calculation formula of the feature extraction model includes: in, This represents the feature vector of the real-time influent components. This represents the full-spectral data of the standard tobacco wastewater. This represents the load matrix obtained through training with historical spectral data.

6. The method for treating tobacco wastewater according to claim 1, characterized in that, The drug delivery parameters include the type of drug and the drug delivery rate. Step S4 includes: Step S41: Match the real-time influent component feature vector with the preset treatment target mode to determine the type of reagent to be added; Step S42: Determine the dosing rate of the reagent based on the predicted influent flow rate, the predicted influent water quality parameters, and the type of reagent to be added.

7. The method for treating tobacco wastewater according to claim 6, characterized in that, Step S4 further includes: Step S43: Based on the production formula code of the standard tobacco production plan data, call the pre-stored formula-component feature spectrum library to generate a predicted component feature vector; Step S44: Calculate the deviation data between the real-time influent component feature vector and the predicted component feature vector, and obtain the deviation correction coefficient based on the deviation data; Step S45: Input the deviation correction coefficient into the incremental proportional-integral-derivative control algorithm to adjust the type of drug and / or the drug dosing rate.

8. The method for treating tobacco wastewater according to claim 7, characterized in that, The incremental proportional-integral-derivative control algorithm is as follows: in, Indicates the current time Correction amount for drug dosing rate. Indicates the current time The deviation between the real-time influent component feature vector and the predicted component feature vector. This represents the deviation value at the previous moment. This represents the deviation value between the two previous moments. Represents the proportionality coefficient. Represents the integral coefficient. This represents the differential coefficient.

9. The method for treating tobacco wastewater according to claim 6, characterized in that, The preset treatment target mode includes a pollutant removal mode, wherein the pollutant removal mode... The agent being administered is polyaluminum chloride, and the mathematical expression for the predictive dose-response model of polyaluminum chloride is as follows: in, The predicted dosing rate of polyaluminum chloride is expressed in L / hour. To predict the inflow rate, the unit is cubic meters per hour. To predict the influent chemical oxygen demand concentration, the unit is mg per liter; The agent being administered is polyacrylamide, and the mathematical expression for the predictive dose-response model of the polyacrylamide is as follows: in, The predicted dosing rate of polyacrylamide is expressed in L / hour. The inflow rate is predicted in cubic meters per hour.

10. The method for treating tobacco wastewater according to claim 6, characterized in that, The preset processing target mode includes a nicotine recovery mode. The agent being added is a calcium-magnesium-aluminum composite flocculant, and the mathematical expression for the predictive dose-response model of the calcium-magnesium-aluminum composite flocculant is as follows: in, The predicted dosing rate of calcium-magnesium-aluminum composite flocculant is expressed in L / hour. To predict the inflow rate, the unit is cubic meters per hour. The nicotine concentration corresponding to the real-time influent component feature vector is expressed in mg per liter. The agent being administered is a micro-sand suspension, and the mathematical expression for the predictive dose-response model of the micro-sand suspension is as follows: in, The predicted dosing rate of the microsand suspension is expressed in L / hour. The inflow rate is predicted in cubic meters per hour.

Images