Process parameter optimization control method and system for continuously producing negative carbon nanometer calcium carbonate under ammonia-calcium system

Abstract

The present application relates to the field of carbon dioxide storage and utilization of nano calcium carbonate, especially to a process parameter optimization control method and system for continuous production of negative carbon nano calcium carbonate in ammonia-calcium system, comprising: collecting multi-source heterogeneous data of gas-liquid-solid three phases in a continuously running nano-micro reactor in real time; after time synchronization and preprocessing of the multi-source heterogeneous data collected in step S1, inputting the data into a semi-empirical prediction model, the semi-empirical prediction model being configured to output at least real-time supersaturation of calcium carbonate in the reaction system and reaction path selectivity evaluation parameters; according to the real-time supersaturation of calcium carbonate in the reaction system and the reaction path selectivity evaluation parameters output by the semi-empirical prediction model, performing process parameter control. The present application has the advantages of high efficiency, strong operability and low implementation difficulty, not only solves the problem of CO2 end-of-pipe emission reduction, but also upgrades the production process of nano calcium, and can realize continuous and stable production of high value-added nano calcium carbonate.

Description

Technical Field

[0001] This invention relates to the field of nano-calcium carbonate and carbon dioxide sequestration and utilization technology, and in particular to a method and system for optimizing and controlling process parameters for the continuous production of negative carbon nano-calcium carbonate in an ammonia-calcium system. Background Technology

[0002] In the industrial production of calcium carbonate, nano-scale preparation represents an important direction for the advancement of materials. Currently, the traditional mainstream batch carbonization process for calcium hydroxide is still widely used. This process typically involves feeding a calcium hydroxide slurry obtained from lime digestion into a reactor, introducing carbon dioxide gas for carbonization, and then unloading the entire batch after the reaction is complete. Although this method has advantages such as flexible operation, low equipment investment threshold, and high tolerance to raw material fluctuations, its inherent problems of large batch stability fluctuations and low energy efficiency have long constrained the improvement of production efficiency and cost competitiveness. Especially against the backdrop of stringent requirements for the morphology, particle size, crystallinity, and distribution of nano-calcium carbonate in fields such as electronics, pharmaceuticals, and high-end coatings, the slow control and scale-up effect of batch processes are becoming increasingly prominent. Batch production makes it difficult to completely reproduce the particle size and morphology of each batch of products, with a large amount of non-production time consumed in the feeding, unloading, and cleaning processes, resulting in low space-time yield. In addition, the calcium hydroxide slurry is in a strongly alkaline environment, and the introduction of carbon dioxide can easily lead to excessively high supersaturation in local areas, triggering explosive nucleation and particle agglomeration, making the controllability of the fine structure of the product a bottleneck.

[0003] To overcome the aforementioned problems, production processes are evolving along two key dimensions: first, a shift from batch to continuous engineering paradigms; and second, a chemical pathway innovation from traditional calcium hydroxide slurry carbonation to ammonia-calcium solution systems. In the ammonia-calcium system, soluble calcium salts and ammonia water are used as raw materials, and the carbonation reaction takes place in a buffered environment of ammonia-ammonium ions. This system utilizes the buffering effect of ammonia to ensure a smooth and controllable pH change, thereby facilitating precise control over the nucleation and growth rates of calcium carbonate, providing a chemical basis for the preparation of nanoparticles with specific morphologies and narrow particle size distributions. Furthermore, this system can directly utilize low-grade ores or industrial by-product calcium salts, demonstrating potential for comprehensive resource utilization. However, the ammonia-calcium system involves NH3 / NH4... + Buffer, Ca 2+ The precise coupling of multiple chemical equilibria, such as complexation and CO2 dissolution and ionization, shifts the control objective from "completing the reaction" to "guiding the reaction path," placing higher demands on the real-time and precise control of key parameters such as pH and supersaturation.

[0004] Combining continuous production with the ammonium calcium system to develop a "gas-liquid-solid three-phase continuous inlet and outlet ammonium calcium system process" faces significant control challenges. Continuous production requires the reaction system to be in a macroscopic steady state of dynamic equilibrium; even minor fluctuations in any parameter can be amplified along the way, directly affecting the quality of the final product. Simultaneously, the system itself possesses the complexity of multi-step ionic reactions and complexation equilibration, further escalating the process control task. It requires precisely coordinating the chemical reaction network coupled with multiple variables such as ammonia concentration, carbon dioxide partial pressure, and ionic strength within a continuous flow, multi-phase mixing physical field, and stabilizing its output within the optimal parameter window for the target product.

[0005] Therefore, existing technologies still lack a precise, efficient, and easy-to-implement control method and strategy suitable for the continuous production process of the ammonium calcium system. The traditional experience-based "observation-adjustment" model is no longer sufficient to meet the system's requirements for real-time sensing, model prediction, and dynamic optimization. There is an urgent need to establish an intelligent control system based on data-driven principles and with feedforward-feedback composite control capabilities to achieve stable control of the continuous production process of nano-calcium carbonate and improve product consistency.

[0006] In view of this, the present invention is proposed. Summary of the Invention

[0007] The purpose of this invention is to provide a method and system for optimizing and controlling process parameters for the continuous production of negative carbon nano-calcium carbonate in an ammonia-calcium system. This control method not only solves the problem of CO2 end-of-pipe emission reduction, but also upgrades the production process of nano-calcium, enabling the continuous and stable production of high-value-added nano-calcium carbonate.

[0008] The first aspect of this invention provides a method for optimizing and controlling process parameters in the continuous production of carbon-negative nano-calcium carbonate in an ammonia-calcium system. This method is applied to the preparation of nano-calcium carbonate using a calcium ion-containing leachate and a carbon dioxide-containing gas in a continuously operating nano-microreactor, and includes the following steps: S1. Real-time acquisition of multi-source heterogeneous data of the gas-liquid-solid three-phase system within a continuously operating nano-microreactor, wherein the data includes at least pH value, Ca... 2+ Concentration, NH3 concentration, NH4 + Concentration, CO3 2- Concentration, HCO3 - Information such as concentration, temperature, bubble size distribution and gas holding capacity, particle size and distribution of products in the system; S2. After time synchronization and preprocessing of the multi-source heterogeneous data collected in step S1, the data is input into a semi-empirical prediction model. The semi-empirical prediction model is a machine learning model, and a physical consistency constraint term based on the chemical equilibrium constant of the ammonia-calcium system is introduced into the loss function used for its training. The semi-empirical prediction model is configured to output at least the real-time supersaturation of calcium carbonate and the reaction path selectivity evaluation parameters in the reaction system. S3. Based on the real-time supersaturation of calcium carbonate and the reaction path selectivity evaluation parameters of the reaction system output by the semi-empirical prediction model in step S2, process parameter control is performed.

[0009] In a preferred embodiment of this technical solution, the data includes key component solution chemistry data, gas-liquid-solid three-phase characteristic data within the reactor, and process operation and state parameter data. Through the key component solution chemistry data, the selectivity of the main reaction and competing reactions, and the magnitude of supersaturation are analyzed. Through the gas-liquid-solid three-phase characteristic data within the reactor, the kinetics of three-phase mass transfer and reaction are analyzed. Through the process operation and state parameter data, the stable operation of the entire system is analyzed. Ultimately, trend prediction is achieved, and operational recommendations are provided.

[0010] In a preferred embodiment of this technical solution, in step S1, the data includes the velocity distribution, particle size distribution, aggregation state, gas holding capacity, and local solid content of bubbles and particles, as well as the pH value and key ion concentration (Ca) in the reaction system. 2+ NH3 / NH4 + CO3 2- HCO3 - The concentration of CO2, the CO2 concentration and flow rate at the reactor inlet and outlet, and the temperature, pressure, and liquid and gas feed flow rates of the reaction system.

[0011] The above parameters, combined with the data and data processing results, can be used to calculate the macroscopic kinetics of reaction crystallization, thereby estimating the crystal growth rate and providing guidance for process control in the production process from a kinetic perspective: when the crystal growth rate is too high, a system warning is given, and a comprehensive judgment is made based on the thermodynamic parameters obtained above, and intervention measures such as pH, temperature, concentration, and flow rate are taken to reduce the reaction rate.

[0012] In principle, based on the dynamic particle size measurement results and using the population balance model, the system can calculate the product particle size growth rate r. G and the fitting-related rate constant (typically 10). -11 ~10 -10 (in the order of magnitude m / s), r G It can also be related to the thermodynamic parameter of supersaturation S (generally r GThese are functions of different forms, so the system can correlate thermodynamics with kinetic empirical formulas; at the same time, all system judgments can be cross-validated from kinetic and thermodynamic perspectives, and kinetic calculations can also preliminarily predict the effect of adjustment operations on the average particle size or particle size distribution that may be achieved.

[0013] More preferably, with the help of signal auto / cross-correlation analysis and reconstruction calculation, the velocity distribution, particle size distribution, and the generation and destruction of bubble / particle agglomerates in the nano-micro reactor channel are collected and monitored in real time and dynamically using optical fiber probes and industrial CT imaging equipment, as well as the changes in local solid content or gas holding capacity in the channel; at the same time, the particle size is monitored online using a dynamic light scattering particle size analyzer.

[0014] More preferably, product liquid is periodically collected at the reactor outlet, the specific surface area of ​​the sample is analyzed offline, and the particle size distribution of the nanosample is analyzed offline using a scanning / transmission electron microscope or a dynamic light scattering analyzer, and compared and verified with the corresponding data of the above online measurement.

[0015] More preferably, ion-selective electrode arrays or online ion chromatography are used to monitor and collect pH and Ca in the reactor in real time. 2+ NH4 + / NH3、CO3 2- HCO3 - Time series of concentration signals of key components or ions.

[0016] More preferably, the concentration and flow rate of CO2 continuously entering and exiting the reactor, as well as the flow rate of the liquid phase feed, are closely monitored by an infrared CO2 analyzer, and accurate records are kept of changes in other state parameters such as temperature and pressure of the reaction system.

[0017] More preferably, the nano-microreactor is a tubular nano-microreactor.

[0018] As a preferred embodiment of this technical solution, the semi-empirical prediction model is a neural network model.

[0019] As a preferred embodiment of this technical solution, the semi-empirical prediction model includes an embedded computational unit for calculating intermediate variables based on the chemical equilibrium equation.

[0020] As a preferred embodiment of this technical solution, in step S2, the variables input to the semi-empirical prediction model include at least the key ion concentration, pH, temperature, bubble size, gas flow rate, and liquid flow rate. In step S3, the variables output by the semi-empirical prediction model include at least calcium carbonate supersaturation, CO2 absorption rate, and Ca2+. 2+ Conversion rate and particle size prediction values.

[0021] In a preferred embodiment of this technical solution, in step S2, the equilibrium constant includes the ionization constant K of water. w The ionization constant K of ammonia b Ammonium ion hydrolysis constant K h CO2 solubility constant K sol The first-order ionization constant K of carbonic acid a1 With the second-order ionization constant K a2 The solubility product constant K of calcium carbonate sp The bicarbonate autoionization constant K' and the main reaction constant for acid-base neutralization K R .

[0022] In a preferred embodiment of this technical solution, in step S2, the reaction pathway selectivity evaluation parameter is the calculated HCO3. - Concentration and CO3 2- The ratio of concentrations; When the ratio exceeds a preset threshold, the NH4 content is adjusted. + Adjust the feed amount and / or lower the system temperature to guide the reaction toward the formation of calcium carbonate as the primary component.

[0023] As a preferred embodiment of this technical solution, in step S3, when adjusting the process parameters, at least one of the following operations is performed: (a) Adjust the amount of ammonia or ammonium salt added to the reaction system, and / or adjust the inlet pressure or flow rate of carbon dioxide gas; (b) Adjust the temperature of the reaction system or set a temperature gradient along the reactor flow channel; (c) Adjust the circulation rate of the liquid phase in the reactor or inject a crystal form control agent; Among them, when the real-time oversaturation exceeds the preset range, the control operation (a) is executed; When the predicted product particle size deviates from the target range, an adjustment operation (c) is performed. As a preferred embodiment of this technical solution, the calcium-containing leaching solution is obtained by leaching high-calcium industrial solid waste, and the carbon dioxide-containing gas is industrial waste gas. The high-calcium industrial solid waste includes any one of high-calcium red mud, carbide slag, high-calcium fly ash, steel slag, and waste concrete. The industrial waste gas includes any one of boiler flue gas, gasifier tail gas, blast furnace tail gas, and cement kiln tail gas.

[0024] A second aspect of the present invention provides a process parameter optimization and control system for the continuous production of carbon-negative nano-calcium carbonate in an ammonia-calcium system, specifically comprising: The data acquisition module is used to collect multi-source heterogeneous data of the gas-liquid-solid three phases in a continuously operating nano-micro reactor in real time. An edge computing and data processing module, deployed in the edge computing unit, is used to receive and process data from the data acquisition module and run the aforementioned semi-empirical prediction model. The optimization decision and control module is used to receive the predicted values ​​output by the edge computing and data processing module, generate control instructions, and send the control instructions to the execution mechanism; The data acquisition module includes at least an online analyzer for monitoring ion concentration, a probe or imaging device for monitoring bubble parameters, and a gas analyzer for monitoring CO2 concentration and flow rate. The actuator includes at least an ammonia or ammonium salt addition device, a gas flow or pressure regulating device, and a temperature regulating device.

[0025] More preferably, the actuator includes an ammonia or ammonium salt metering pump, a CO2 inlet pressure regulating valve, a liquid phase circulation pump, a crystal form control agent injection device, and a temperature regulating device.

[0026] More preferably, the edge computing unit employs an embedded system or a microprocessor cluster and integrates a machine learning acceleration unit.

[0027] A third aspect of the present invention provides a negative carbon nano-calcium carbonate product prepared by optimizing and controlling the process parameters of the continuous production of negative carbon nano-calcium carbonate in the above-mentioned ammonia-calcium system.

[0028] For example, in a 10 kg / h process using high-calcium red mud-power plant flue gas as input... - ¹In a continuous production unit for carbon-negative nano-calcium carbonate, an average particle size of 50-100 nm and a specific surface area of ​​20-35 m² can be obtained. 2 / g, and the net CO2 sequestration per ton of product is 0.14 tons; In the mineralization of low-concentration CO2 using calcium carbide slag leachate, spindle-shaped nano-calcium carbonate with a purity higher than 98 mol% and uniform morphology can be obtained, while achieving a carbon sequestration efficiency of about 0.4 tons of CO2 / ton of product. In treating boiler flue gas with high-calcium fly ash leachate, an average particle size of 80 nm and a concentrated particle size distribution (D) can be obtained. 90 / D 10 <2.5 or D 50 High-quality nano calcium carbonate with / D[4,3]≈1); In the mineralization of blast furnace tail gas containing CO2 as negative carbon nano-calcium carbonate in steel slag leaching solution, a product with a stable specific surface area of ​​25±2 m² can be obtained. 2 Negative carbon nanofiber calcium carbonate in the range of / g.

[0029] The method for optimizing and controlling process parameters for the continuous production of carbon-negative nano-calcium carbonate in an ammonia-calcium system of the present invention has at least the following beneficial effects: In the process parameter optimization and control method for the continuous production of negative carbon nano-calcium carbonate in the ammonia-calcium system of this invention, the leachate is introduced into the tube-side flow channel of a tubular nano-calcium reactor and comes into efficient contact with CO2 nanobubbles diffused from the shell side of the reactor into the tube side to generate negative carbon nano-calcium carbonate, which then leaves the reactor in a timely manner. The core of the continuous production process of this invention lies in the integration of a semi-empirical prediction model of fluid mechanics, rheology, and solution chemistry. It considers secondary mining data such as the three-phase mass transfer area and mass transfer coefficient, the ion product of different ions, thermodynamic and macroscopic kinetic parameters calculated from direct detection data. Based on the secondary mining data, the supersaturation is calculated, the degree of progress of the main chemical reactions in the solution system and the trend of particle size change are predicted, and the operation is guided by the addition amount of gaseous, liquid and solid raw materials and the adjustment of the system reagent regime.

[0030] Therefore, in a gas-liquid-solid three-phase highly dispersed nano-micro system, the process parameter optimization and control method for continuous production of negative carbon nano-calcium carbonate under the ammonia-calcium system of this invention is adopted. By using the calcium ion-containing leachate of high-calcium industrial waste residues (such as high-calcium red mud, high-calcium fly ash, carbide slag, waste concrete or steel slag, etc.) to mineralize CO2 (the gas source often comes from industrial waste gas of different concentrations, such as boiler flue gas, gasifier tail gas, blast furnace tail gas, etc.), and by precisely controlling the supersaturation of calcium carbonate, the particle size distribution and bubble dispersion of nano-micro bubbles in the gas-liquid-solid three-phase system, the continuous and stable production of high-value-added nano-calcium carbonate can be achieved. Attached Figure Description

[0031] To more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the specific embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0032] Figure 1 This is a flowchart illustrating the principle of the process parameter optimization and control method for the continuous production of negative carbon nano-calcium carbonate in the ammonia-calcium system of the present invention. Figure 2 This is a detailed stage division information analysis diagram of the process parameter optimization and control method for continuous production of negative carbon nano-calcium carbonate in the ammonia-calcium system of the present invention; Figure 3 This is a comparison image of the morphology of samples under different working conditions in the continuous test of Example 3 of the present invention after drying, under a transmission electron microscope. Figure 4 This is a SEM image of the product obtained in Example 4 of the present invention. Detailed Implementation

[0033] It should be noted that the following detailed descriptions are illustrative and intended to provide further explanation of this application. Unless otherwise specified, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application pertains.

[0034] It should be noted that the terminology used herein is for the purpose of describing particular implementations only and is not intended to limit the exemplary implementations according to this application. As used herein, the singular form includes the plural form unless the context clearly indicates otherwise. Furthermore, it should be understood that when the terms "comprising" and / or "including" are used in this description, they indicate the presence of features, steps, operations, devices, components, and / or combinations thereof.

[0035] The technical solution of the present invention will be clearly and completely described below with reference to the embodiments. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0036] Example 1 The continuous production of nano-calcium carbonate places significantly higher demands on process parameter monitoring and timely intervention than existing batch-operation mainstream processes, and the calcium carbonate production process in the ammonia-calcium system involves NH4. + CO3 2- and HCO3 - The processes of double hydrolysis of three ions under alkaline conditions (NH4+) + + HCO3 - / CO3 2- +H2O NH3·H2O + H2CO3 / HCO3 - This presents a problem where the solution chemistry of calcium carbonate is far more complex than that of existing mainstream Ca(OH)₂ production processes. This invention proposes a method for optimizing and controlling process parameters in the preparation of nano-calcium carbonate using a calcium ion-containing leachate and a carbon dioxide-containing gas in a continuously operating nano-microreactor. This process requires very strict dynamic monitoring of bubble phase parameters, demanding accurate prediction of bubble diameter distribution and bubble lifespan within the reactor to ensure reaction rate and product particle size.

[0037] like Figure 1-2 As shown, this embodiment provides a method for optimizing and controlling process parameters for the continuous production of negative carbon nano-calcium carbonate in an ammonia-calcium system, specifically including the following steps: S1. Real-time acquisition of multi-source heterogeneous data of gas-liquid-solid three phases in a continuously operating nano-micro reactor. The data includes key component solution chemical data, gas-liquid-solid three-phase characteristic data in the reactor, and process operation and state parameter data. This is to accurately detect and control the core parameters affecting the continuous mineralization process. The system integrates analytical instruments with different sampling frequencies and principles, and develops a centralized, coordinated, and synchronous signal processing method based on the principle of parallelism to integrate various types of data.

[0038] Specifically, it includes: Coupled with a fiber optic probe and a small industrial CT imaging device, the velocity distribution, particle size distribution, and formation and extinction of bubble / particle agglomerates of particles and bubbles in the reactor channel are acquired in real time and dynamically monitored with the help of signal auto / cross-correlation analysis and reconstruction calculations, as well as the changes in local solid content or gas holding capacity in the channel; at the same time, a dynamic light scattering particle size analyzer is used to monitor particle size online. Product liquid was periodically collected at the reactor outlet, and the specific surface area of ​​the samples was analyzed offline. The particle size distribution of the nano samples was analyzed offline using a scanning / transmission electron microscope or a dynamic light scattering analyzer, and the data were compared and verified with the corresponding online measurement data. Real-time monitoring and acquisition of pH and Ca in the reactor using ion-selective electrode arrays or online ion chromatography. 2+ NH4 + / NH3、CO3 2- HCO3 - Time series of concentration signals of key components or ions; The concentration and flow rate of CO2 continuously entering and exiting the reactor are closely monitored using an infrared CO2 analyzer, as well as the flow rate of the liquid feed. Accurate records are also kept of changes in other state parameters of the reaction system, such as temperature and pressure. At the same time, auxiliary cross-validation data is provided for the dataset through periodic sampling and offline analysis.

[0039] S2. After time synchronization and preprocessing of the multi-source heterogeneous data collected in step S1, the data is input into the semi-empirical prediction model. The semi-empirical prediction model is a machine learning model, and a physical consistency constraint term based on the chemical equilibrium constant of the ammonia-calcium system is introduced into the loss function used for its training. The semi-empirical prediction model is configured to output at least the real-time supersaturation of calcium carbonate and the reaction path selectivity evaluation parameters in the reaction system. Because different analytical instruments are based on different physical / chemical principles, their output signal types (concentration, peak area, absorbance, current, etc.), dimensions, and ranges vary greatly, and their data formats and protocols are not uniform. For example, communication protocols and data formats may differ between manufacturers, requiring additional conversion and integration. Furthermore, data from fast-response instruments (such as fiber optics) and slow-response instruments (such as chromatography) need to be aligned through interpolation, resampling, and other methods. To achieve synchronous comparison of different data and provide timely decision-making suggestions, this invention employs a data fusion algorithm to handle data noise and asynchrony, combines soft sensing and machine learning to compensate for errors, and ensures real-time performance through edge computing. Simultaneously, a regular automatic calibration and anomaly diagnosis mechanism is established to improve the overall reliability of the system.

[0040] Specifically, this invention provides a detailed preprocessing method for data from multi-source heterogeneous analytical instruments to achieve standardization, synchronization, and integration of data of different types, dimensions, ranges, and sampling rates. The steps are as follows: ① Data format and communication protocol conversion and integration Data output from instruments from various manufacturers is formatted uniformly through standardized interface adapters. First, for different communication protocols (such as OPC UA, HTTP / RS-232, etc.), a protocol parsing middleware is used for real-time data extraction and conversion into a unified data format. For unstructured or special-format data, a customized data parsing module maps it into a structured data table. Simultaneously, a normalization model for dimensions and ranges is established, and the original signals (such as peak area, absorbance, and current) are uniformly converted based on calibration curves or transformation formulas. This process is completed in real-time at edge computing nodes, ensuring low latency.

[0041] ② Synchronization and alignment method for multi-rate data For fast-response instruments (such as fiber optic sensors with sampling rates >10 kHz) and slow-response instruments (such as chromatographs with sampling rates <0.1 Hz), an adaptive time window alignment strategy is adopted, for example: Interpolation processing: For slow signals, spline interpolation (such as cubic splines) or linear interpolation is performed on the high-speed sampling timestamp to generate equally spaced sequences; at the same time, high-frequency noise is suppressed by moving average filtering; Resampling alignment: A unified time base is set, with the highest sampling rate as a reference, and forward padding or linear resampling is used to fill missing points for low-speed data; for ultra-high-speed data, downsampling is performed while preserving features (such as taking the average value every N points); Event trigger synchronization: For sudden measurement events (such as the appearance of chromatographic peaks), the data segments of associated instruments are aligned according to the event trigger time through timestamp matching and buffer queues.

[0042] ③ Data quality enhancement and anomaly preprocessing techniques High-frequency random noise is eliminated using wavelet denoising. For signal drift, a sliding window difference method is used to detect baseline drift and correct it in real time. In addition, an outlier detection model is established based on historical data distribution to automatically identify and remove outliers, avoiding fusion distortion.

[0043] ④ Real-time edge computing framework The above preprocessing process is deployed on the edge gateway, encapsulating each processing module and executing data access, parsing, and alignment in parallel through a pipeline architecture to ensure response.

[0044] Therefore, this invention achieves efficient integration and real-time alignment of multi-source instrument data through standardized interfaces, adaptive synchronization algorithms, and edge computing, providing high-quality input for subsequent fusion analysis.

[0045] The method for establishing the semi-empirical prediction model in this embodiment is as follows: Using the multi-source heterogeneous data collected in step S1, such as solution ion concentration, macroscopic state parameters of the gas-liquid-solid three-phase system, and characteristic parameters of the bubble and particle phases, a semi-empirical mathematical model is established. This model includes parameters such as pH value, key ion concentration, and bubble and particle size, and indices such as calcium carbonate supersaturation, mass transfer rate, and predicted bubble and particle size. The model is trained and tested using empirical data and laboratory experimental data (such as neural networks, decision trees, and logistic regression). Data mining algorithms are then used to obtain the system's core control variables (i.e., secondary data) and their reasonable variation ranges through theoretical or semi-empirical calculations. The specific process is as follows: ① By combining the collected bubble diameter data with the channel gas holding capacity data, the gas-liquid mass transfer area of ​​the reaction system can be reasonably estimated; while the Ca in the solution entering and leaving the system 2+ Changes in CO2 concentration in the liquid and gas phases can provide conversion and absorption rate data and trends from both the liquid and gas phases, respectively. Furthermore, dynamic changes in the particle size distribution of solid particles can provide information on the stability of particle crystal growth within the system. Based on these data, the control system can calculate the gas-liquid mass transfer kinetics, macroscopic kinetics of the mineralization reaction, and nucleation kinetics of calcium carbonate crystallization. ② Ca obtained through an online ion concentration acquisition system 2+ With CO3 2- The concentration, and when necessary, offline monitoring, can be used to monitor the supersaturation of unprecipitated CaCO3 seed crystals in the reactor solution in real time. This value has a significant impact on particle size distribution. The introduction of this algorithm into process control can provide timely particle size warning information to on-site engineers. Through ① and ②, the system finally establishes the supersaturation S ( The differential correlation between dL / dt and the characteristic crystal grain size L is dL / dt = k S βThis allows for the control of particle size distribution by adjusting thermodynamic state parameters p, V, T, and component concentration, and enables the comparison and prediction of potential control effects, leading to appropriate selections.

[0046] ③ Collecting temperature data can estimate the CO2 solubility under operating conditions and extrapolate HCO3. - Possible trends (excess HCO3) - This will severely affect the selectivity of the reaction system for CaCO3 products); and simultaneously, it will affect the pH and Ca... 2+ NH4 + / NH3、CO3 2- HCO3 - Concentration change trend data can be used to analyze the dynamic equilibrium shift of chemical processes such as hydrolysis / ionization / mineralization within the system and its impact on the selectivity of the above reactions, and can provide timely early warning information or retrospective analysis of process problems that have occurred. ④ Compare and perform correlation analysis on the changing trends of the secondary data obtained above with production indicators such as product purity and particle size to obtain the model prototype. At the same time, adjust the details of the model parameters through laboratory experiments.

[0047] In this step, the existing production and pilot-scale data are first used, and laboratory experiments are conducted to refine the data logic. The predicted values ​​are then matched with actual values ​​under various complex operating conditions to achieve automatic optimization and decision-making of the algorithm, resulting in a reliable optimization and adjustment method. Specifically, the optimization method is cross-trained and validated using training and testing data obtained from real-world application scenarios. This accurately predicts process influencing factors while significantly reducing the interference of uncertainties, providing a basis for process adjustment and even design.

[0048] The method of cross-training and validating the optimization method using training and testing data obtained from real-world application scenarios specifically includes the following steps: Historical production data and laboratory test data were selected for training the model; ① Standard steady-state operating conditions: Standard production condition data corresponding to different product specifications (e.g., different particle sizes); ② Non-steady-state operating conditions (e.g., start-up and shutdown): Raw material fluctuations such as changes in calcium source concentration, CO2 purity, or inlet pressure and flow rate within a certain range. Planned adjustments to key parameters such as step / continuous changes, pH setpoint, and reaction temperature; ③ Abnormal or accidental operating conditions: Simulated scenarios such as scaling, clogging, and metering pump deviation in the laboratory or pilot plant; ④ Laboratory-enhanced test data: Factorial design or response surface optimization test data conducted in laboratory reactors for extreme or high-risk parameter combinations that are difficult to access in production (e.g., extremely high supersaturation, runaway temperature, or temperature loss). Actively explore and establish input-output relationships within a wider parameter range.

[0049] Considering the need for large amounts of data for models such as neural networks, and taking into account the operational realities of different indicator scenarios, the initial training sample size should be no less than 500-1500 sets. In subsequent pilot-scale testing, to achieve robust predictions and effectively reduce uncertainty, the total sample size should ideally reach 10,000-20,000 sets or more. The samples should cover as many of the aforementioned operating conditions as possible. Furthermore, each set of valid data should be collected at the same time or within the same stable period, and all input and output values ​​should be synchronously collected into the data vector.

[0050] Each valid sample contains all input variables (such as pH, Ca). 2+ The values ​​are measured or reliably calculated values ​​(such as concentration, temperature, bubble diameter, etc.). For a very small number of variables that cannot be measured online temporarily (such as specific ion concentrations), data can be filled through high-frequency sampling analysis in the laboratory or through validated reliable online models to ensure the integrity of the sample.

[0051] Each sample corresponds to a real, accurate measured output value, which serves as a benchmark for model learning and validation. Key outputs (such as the average particle size and distribution of the actual product, actual CO2 absorption, calcium conversion rate, etc.) are also analyzed and cross-validated against samples collected concurrently using offline standard instruments (such as ion chromatography, electron microscopy, and titration). For example, the actual supersaturation needs to be calculated based on the ion concentration and temperature obtained from offline titration of the reaction solution.

[0052] The machine learning algorithm with embedded physical constraints is used for training; This application constructs a high-quality dataset with broad coverage, sufficient samples, and complete and real data, and trains it using a machine learning algorithm with embedded physical constraints. This ensures that the obtained optimization model has high prediction accuracy and clear physical meaning, thereby providing a reliable basis for the optimization, control, and design of industrial processes.

[0053] In this invention, no specific algorithm is limited. Conventional basic model algorithms (such as gradient boosting algorithm) can be combined with optimization algorithms (such as Bayesian algorithm), depending on the specific situation.

[0054] like Figure 2 As shown, the input layer of the semi-empirical prediction model of this invention contains n nodes, which correspond to n key process parameters collected in real time, such as: pH value in the reactor, Ca²⁺, etc. + Ion concentration, NH4 + Data includes ion concentration, reaction temperature, CO2 concentration, average bubble diameter and distribution density, particle size and velocity, and reactant / product flow rates. Each node can normalize the received data.

[0055] The semi-empirical prediction model of this invention contains at least one hidden layer, and the number of nodes in each layer can be adjusted according to the data complexity (e.g., 8, 12, or 256 nodes). The hidden layer is used to learn and characterize the highly nonlinear intrinsic relationship between input parameters and complex reaction kinetics (such as nucleation, growth, and mass transfer), and also contains fundamental data such as important reaction thermodynamics and kinetic parameters in the system and gas-solid phase solubility data.

[0056] The output layers of the semi-empirical prediction model of this invention correspond to the key process indicators that need to be predicted. Typical outputs include: calcium carbonate supersaturation (SI = [Ca...]). 2+ CO3 2- ] / K sp ), CO2 absorption rate, Ca 2+ Conversion rate and predicted average particle size or particle size distribution of the product, etc.

[0057] The semi-empirical prediction model of this invention employs a fully connected approach between adjacent layers. Each input node is connected to all hidden layer nodes, and each node in the preceding hidden layer is connected to all nodes in the output layer (or the following hidden layer). The control network learns and adjusts the weights and biases of these connections to establish a high-dimensional mapping function from multi-dimensional inputs to multi-dimensional outputs, which can be integrated to form the transfer function of the control process.

[0058] In the semi-empirical prediction model of this invention, physical and chemical laws are not merely background knowledge, but are deeply embedded in the model training and structure in a collaborative manner with data information, serving as both constraints and features.

[0059] For example, as a physical constraint in the model output: the calculation of the calcium carbonate supersaturation node in the output layer directly incorporates physical formulas. The model does not directly predict this value, but rather achieves it through a customized output module: the model outputs three intermediate variables: system temperature T, Ca... 2+ and CO3 2- Concentrations (named A, B, and C respectively), then calculated using the built-in formula f(B)f(C) / K sp (A) Perform supersaturation calculation, where K sp (T) is the calcium carbonate solubility product constant determined in real time through a built-in database based on the temperature node values ​​in the current input layer. This ensures that the supersaturation prediction follows thermodynamic equilibrium constraints. Furthermore, the next step will be to predict nucleus growth based on a built-in empirical correlation between supersaturation and calcium carbonate growth rate, and compare this prediction with real-time data from the particle size measurement node to indicate the direction of particle size change.

[0060] For example, as a guide for feature engineering, theoretical knowledge guides the selection and construction of input features. For instance, the input may include not only the original concentration, but also ionic strength and ionic activity calculated based on reaction equilibrium and actual analytical concentrations, and CO32 calculated based on pH and equilibrium constants. 2- Derivative features such as pre-estimated concentration values ​​provide more direct physical correlation information for process control.

[0061] Furthermore, during model training, the loss function not only includes the standard error between predicted metrics (such as CO2 absorption rate, particle size, and conversion rate) and actual measurements, but can also include physical consistency loss. For example, based on chemical reaction equations and material balance, a loss function for Ca can be constructed. 2+ The soft constraint on the relationship between consumption and CO2 absorption is incorporated as a penalty term into the total loss function. This makes the model more physically plausible and interpretable in its predictions of complex outcomes, while simultaneously fitting the data.

[0062] The reliability evaluation indicators of the model in this embodiment include correlation coefficient, standard mean square deviation, etc., as detailed in Table 1.

[0063] Table 1

[0064] y i For the observed values, i For predicted values, The mean of the predicted values ​​is x, where n is the number of samples and x is the mean of the predicted values. i For a certain factor Based on the above technical solution, and further preferably, for the model evaluation of complex solution-phase chemical dynamic processes (such as reactors and control systems) in the ammonia-calcium-carbon system, the integral squared error ISE can also be introduced. Or the integral absolute error IAE = When it is necessary to emphasize the penalty for later-stage errors, the integral time absolute error ITAE can be integrated. When necessary, the time-domain and frequency-domain response matching calculations will be invoked to compare the similarity between the model and the step response and frequency response of the actual system.

[0065] As can be seen, the semi-empirical model of this invention belongs to the gray box model. Through multi-level and deep physical and chemical embedding, it realizes that the prediction results are constrained by basic laws such as thermodynamics and mass conservation (reliability); the physicochemical equations provide a reliable inference framework (extension); and the predictions are all supported by corresponding theoretical analysis. It can not only predict, but also explain, and propose how to adjust based on the underlying principles (relatively transparent decision-making). That is, the "principle embedding" of this invention is not a simple data preprocessing or post-processing, but a fusion of the entire process from model internal structure design and training to output interpretation, which constitutes the core feature of "semi-empirical" with a certain degree of chemical intelligence, which is different from pure data-driven methods.

[0066] The specific process of chemical intercalation is as follows: First, a robust chemical calculation kernel is constructed. The model's internal structure doesn't merely learn a point-to-point mapping from input to output. In fact, it includes a "chemical equilibrium calculation module" and a "crystallization kinetics (nucleation growth) calculation module" based on the aforementioned series of chemical reaction equations and thermodynamic / kinetic principles. This module uses real-time acquired [Ca...]... 2+ ]、[NH4 + ], pH (or [H + Using factors such as temperature as core inputs, the system simultaneously solves the kinetic equations of the mass action law, the crystal nucleation equation, the charge balance equation, and the mass conservation equation (and separately examines the proton conservation equation when necessary). This computational kernel can calculate in real time the concentration and supersaturation of key intermediate species that cannot be directly measured but are crucial for prediction. This ensures that the model's description of the system's solution chemical state does not solely rely on data correlation, preventing the model from outputting thermodynamically impossible results. The specific calculation processes described above can all be implemented in automatic control systems using conventional dynamic characteristic description methods (such as differential equation description of components, transfer functions, response curve-assisted judgment, equivalent transformation of components, etc.). Secondly, chemical principles are also integrated as a supervisor and corrector for model training and optimization. For example, a so-called physical consistency penalty can be introduced through a customized loss function: during the model training phase, the loss function not only includes error terms for prediction and measurement, but can also add material balance residuals: calculating the closure error based on the input total calcium and total carbonate (derived from CO2 absorption and calcium conversion) and the output precipitation amount and residual ion concentration; or adding charge balance residuals: calculating whether the predicted ion concentrations meet the electroneutrality condition of the solution; and reaction rate correlation residuals: based on the macroscopic kinetic model, correlating the CO2 absorption rate (mass transfer on the gas phase side) with the carbon dioxide formation rate in the liquid phase. This improves the model's extrapolation capability.

[0067] Finally, the underlying principles can be embedded into decision-making control. This involves linking model outputs to interpretable chemical state indicators such as supersaturation, CO2 uptake efficiency, and Ca2+.2+ Conversion rate, by its very nature, is an indicator with clear physical meaning, directly calculated or derived from the internal chemical calculation kernel. This allows operators to directly understand the basis of the model's predictions. For example, if the model predicts that the product particle size will become coarser, it will also provide an explanation such as "the current supersaturation decreases due to the increase in temperature, resulting in a decrease in nucleation rate," avoiding simply providing a black-box suggestion or conclusion.

[0068] It should also be noted that the semi-empirical gray-box model can also be used to construct an "influence factor analysis" module based on chemical competitive reactions (such as...). Figure 2 (As shown). For example, after model inference, a sensitivity analysis or factorial module can be added. By fine-tuning the inputs (such as pH and temperature), the changing trends of the concentrations of key intermediate species and competing reaction equilibria (such as ammonia ionization vs. bicarbonate hydrolysis) can be observed, thereby quantitatively assessing the rise and fall of different reaction pathways, further explaining why yields or selectivity change under certain operating conditions, and making targeted recommendations (such as "It is recommended to fine-tune the pH to suppress ammonium NH4+"). + Hydrolysis, thereby protecting CO3 2- "Supply", to achieve a closed loop from forecasting to analysis to recommendations.

[0069] In summary, the semi-empirical prediction model of this invention is specifically manifested in a computer as an embedded physical and chemical information neural network program structure, or a hybrid / integrated program structure based on theoretical constraints. Essentially, it is an integrated prediction and decision-making program specifically designed for the CO2 mineralization system of the ammonia / calcium system, deeply coupling mechanistic models such as industrial crystallization kinetics, solution chemistry, and thermodynamics with various machine learning algorithms. The program encapsulates mechanistic equations (such as crystallization nucleation and growth kinetic equations, phase equilibrium calculations, and supersaturation prediction models) and basic data at the underlying level, embedding them as prior constraints within a machine learning framework. It also integrates classification / regression algorithms such as logistic regression, decision trees, and support vector machines, as well as unsupervised learning algorithms such as clustering and dimensionality reduction. Through an ensemble learning strategy, the outputs of each model are combined to form a hybrid model that combines mechanistic rationality with data fitting capabilities. The model has a standard input layer, hidden layer, and output layer, but it embeds a computational unit written according to chemical equilibrium equations (such as the law of mass action and charge balance equation) at a specific location within it. This unit receives intermediate variables from the network and outputs results that strictly follow physical laws (such as supersaturation and reaction entropy). At the same time, its computation process can participate in the backpropagation training of the entire network.

[0070] Among these algorithms, logistic regression, decision trees, and clustering are not integral parts of the core model itself, but rather supporting tools that work in conjunction with the core neural network throughout the system development and data preprocessing stages. For example, dimensionality reduction and clustering algorithms are used for pattern recognition and feature selection from historical data; white-box models such as decision trees are used for preliminary feature importance analysis. In the final hardware and software system, these algorithms may also constitute auxiliary or rule-based modules, providing verification, early warning, or interpretability supplements for real-time predictions. In short, various machine learning algorithms are methods and tools used in the construction and optimization of the core model to jointly achieve the accuracy and reliability of industrial process predictions.

[0071] S3. Based on the real-time supersaturation of calcium carbonate and the reaction path selectivity evaluation parameters of the reaction system output by the semi-empirical prediction model in step S2, process parameter control is performed. When adjusting process parameters, at least one of the following operations must be performed: (a) Adjust the amount of ammonia or ammonium salt added to the reaction system, and / or adjust the inlet pressure or flow rate of carbon dioxide gas; (b) Adjust the temperature of the reaction system or set a temperature gradient along the reactor flow channel; (c) Adjust the circulation rate of the liquid phase in the reactor or inject a crystal form control agent; When the real-time oversaturation exceeds the preset range, control operation (a) is executed; When the predicted product particle size deviates from the target range, an adjustment operation (c) is performed. The reaction pathway selectivity evaluation parameter is the calculated HCO3. - Concentration and CO3 2- The ratio of concentrations; When the ratio exceeds a preset threshold, the NH4 content is adjusted. + Adjust the feed amount and / or lower the system temperature to guide the reaction toward the formation of calcium carbonate as the primary component.

[0072] Based on the completion of data acquisition and reliable data analysis, the embedded model of the process system, combined with existing experience data, can make predictions of particle size distribution, conversion rate, and absorption rate of the system with a confidence level of over 80%. Based on the prediction results, scenario analysis is conducted, and suggestions for adjusting the operating points of the reactor and equipment are provided. Furthermore, new production and experiments can be organized to cross-validate the model's prediction and control effects, further refining the control and optimization methods.

[0073] Specifically, the following steps are included: The supersaturation of calcium carbonate under different conditions was calculated by relating key ion concentrations, pH values, CO2 partial pressures and concentrations to the product constant of calcium carbonate ions in the system. The particle size change status of the product is obtained from online / offline information of particle size distribution. Combined with the aforementioned supersaturation data and online bubble size and dispersion data, the differential and integral trends of particle size distribution change are predicted by crystallization kinetics and chemical kinetics calculations. By calculating the solubility products of different primary and secondary reactions in the solution system using various key ion concentrations, and combining them with their corresponding equilibrium constants, the competitiveness of each reaction in the reactor is compared and evaluated. The selectivity of the solution chemical reaction system for the calcium carbonate main product is predicted, and corresponding suggestions are proposed, along with the resulting positive and negative impacts. For example, it is necessary to comprehensively consider the trends of change between the following chemical reactions, taking into account temperature, flow rate, particle size, and gas-liquid dispersion, to determine what beneficial or detrimental process changes will occur in the reactor at different temperatures and pH values: The ionization of water (H2O) H + + OH - K w =10 -14 The ionization of ammonia (NH3·H2O) NH4 + + OH - K b =1.8×10 -5 Ammonium ions hydrolyze NH4 + + H2O NH3·H2O + H + K h =5.6×10 -10 CO2 dissolves CO2 + H2O H2CO3, K sol = 3.3 × 10 -2 Carbonic acid ionization H2CO3 H + + HCO3 - K a 1 = 4.2 × 10 -7 Bicarbonate ionization HCO3 - H + + CO3 2- K a 2 = 5.6 × 10 -11 carbonate ion hydrolysis CO3 2- + H2O HCO3 - + OH - K h =1.8×10-4 Bicarbonate ions hydrolyze HCO3 - + H2O H2CO3+ OH - K h =2.4×10 -4 Bicarbonate self-ionization HCO3 - + HCO3 - H2CO3+ CO3 2- K' = 1.3 × 10 -4 Calcium carbonate precipitation-dissolution equilibrium Ca 2+ + CO3 2- CaCO3↓, K sp = 2.8 × 10 -9 The main acid-base neutralization reaction is NH3·H2O + CO2 NH4 + + HCO3 - K R =7.6×10 2 ; Combined with Ca 2+ And CO2 concentration data to estimate Ca from the liquid and solid phases 2+ The conversion rate and CO2 absorption rate are determined, and the macroscopic stability of the reactor and its working / operation point are determined based on macroscopic data such as reactor temperature, pressure, and yield. This helps process operators balance and make trade-offs between product quality (such as particle size and crystal form) and product yield.

[0074] Therefore, the core of the process parameter optimization and control method for the continuous production of negative carbon nano-calcium carbonate in the ammonia-calcium system of the present invention is as follows: in an environment that may have both chemical and electrochemical corrosion, the main monitoring parameters, including key ion content, pH value, CO2 supply, bubble and particle parameters, reaction temperature and additive dosage, are obtained by integrating the physical and chemical parameter signals of the ammonia / calcium carbonation double hydrolysis reaction system. The physicochemical data are then further analyzed to provide suggestions on the underlying judgment mechanism and logic for the operation layer from the perspectives of fluid dynamics, solution chemistry, thermodynamics-kinetics and specific process details.

[0075] Example 2 This embodiment provides a process parameter optimization and control system for the continuous production of negative carbon nano-calcium carbonate in an ammonia-calcium system, specifically including: The data acquisition module is used to acquire multi-source heterogeneous data of the gas-liquid-solid three phases in a continuously operating nano-micro reactor in real time. Specifically, it includes optical fiber probe array, industrial CT imager, online ion chromatograph, ion selective electrode array, infrared CO2 analyzer, dynamic light scattering particle size analyzer, temperature sensor, pressure sensor and flow meter, etc. An edge computing and data processing module, deployed in the edge computing unit, is used to receive signals from the data acquisition module and perform data fusion, noise filtering and timestamp alignment, as well as run a semi-empirical prediction model. In terms of program structure, this module includes a data preprocessing submodule, a mechanism calculation submodule based on chemical equilibrium and kinetic parameters, and a machine learning model training and prediction submodule. Each submodule is coupled through a unified data interface and parameter passing mechanism to form an integrated algorithm system. The optimization decision and control module is used to receive the predicted values ​​output by the edge computing and data processing module, generate control instructions, and send the control instructions to the execution mechanism; The actuators include an ammonia or ammonium salt metering pump, a CO2 inlet pressure regulating valve, a liquid phase circulation pump, a crystal form control agent injection device, and a temperature control device.

[0076] The integrated algorithm system is implemented as an executable program, algorithm library, or algorithm plug-in integrated into the industrial control system. It can receive process input variables from the data acquisition module in real time, call the mechanism calculation submodule to calculate theoretical values, and combine the output of the machine learning model submodule to make comprehensive predictions and decisions. At the same time, it supports updating and optimizing the model parameters using new data generated in the production process.

[0077] Example 3 This embodiment uses high-calcium red mud-power plant flue gas as input at a rate of 10 kg / h. - ¹Continuous production unit for negative carbon nano-calcium carbonate.

[0078] In this embodiment, the solid waste required for CO2 mineralization comes from high-calcium red mud (CaO 42.78 wt%, D90=68 µm, moisture content 18 wt%) from an alumina plant in Shandong; the CO2 gas source is power plant flue gas (CO2 14 vol%, SO2 <80 mg / Nm³). 3 The initial temperature was 52℃ and the flow rate was 300 Nm³. 3 / hr).

[0079] High-calcium red mud was first prepared into a slurry at a liquid-to-solid ratio of 3:1, and then 0.3 mol / L of NH4Cl co-solvent was added. 2+ The leaching rate was 92%. After plate and frame filtration, Ca... 2+The concentration of the active ingredient is 0.78 mol / L, pH=10.58; impurities such as Fe, Al, and Mg are below 5 ppmw. The material enters a subsequent continuous mineralization tubular nanoreactor (tube inner diameter 12 mm, effective length 3 m, 120 tubes connected in parallel), with CO2 flowing through the shell side. The designed gas-liquid ratio is 8:1. Based on the characteristics of the reactor and operating conditions, this system requires the use of ion activity... γ i (lg γ i = 0.509 f (z i , I , ), I Z represents ionic strength. i The supersaturation S was calculated using the ion charge, and then S was compared with the particle growth rate r obtained from the particle size distribution system. G By relating them, we obtain the thermodynamic-kinetic joint differential model. k (S-1) 1.5 ln 0.77 (S), used for production control and predictive analysis.

[0080] Online detection nodes: ① Coupled optical fiber probe with industrial CT imaging equipment: bubble / particle concentration is collected once per second to calculate gas holding capacity and solid content; the velocity and diameter distribution of particles and bubbles are measured using a velocimetric fiber probe, with a sampling period of less than 1 second; ② Monitoring pH and Ca values ​​using an ion-selective electrode array and online ion chromatography. 2+ NH3 / NH4 + HCO3 - Concentration, response time < 5 s; ③ The CO2 concentration at both the inlet and outlet is monitored using an infrared CO2 analyzer, with an accuracy of ±1%. ④ The particle size distribution, median particle size and volume average particle size in the reactor are dynamically monitored using a DLS online testing instrument.

[0081] The edge computing unit uses a Raspberry Pi cluster with an embedded Python-TPU for real-time soft measurement and model inference.

[0082] After initial liquid filling of the tube side of the mineralization unit, CO2 is introduced into the shell side, and the system pressure is controlled at 0.25 MPa. The production process adopts gradient feeding: set according to "ammonia / calcium molar ratio 2.2, carbon / calcium molar ratio 1.05"; the volumetric pump controls the liquid phase flow rate at 12.0-12.5 L / h and the gas flow rate at 96-105 L / h.

[0083] The model self-tuning process is as follows: ① After the system started running for 30 minutes, 1800 sets of data were collected; the semi-empirical model performed online regression, and the oversaturation prediction error RMSE was 5.8%, reaching the "excellent" level; ② Particle size drift warning: After 3 hours of operation, the DLS showed that the median particle size D50 increased from 65 nm to 82 nm; the model diagnosed "supersaturation decreased by 12%". After extrapolating based on the existing total ammonia and ammonium concentration data, the system immediately replenished ammonia (4% of the existing concentration level) and simultaneously increased the CO2 feed pressure by 0.02 MPa. After 30 minutes, the median particle size returned to around 68 nm. ③ Composition fluctuations: The CaO content of the red mud feed on the dissolution side suddenly dropped to 37.58%, and half an hour later, the CaO content on the mineralization side... 2+ The concentration was reduced to 0.65 mol / L; the system automatically extended the residence time by 8% through feedforward compensation to ensure that the carbon conversion rate remained above 92%.

[0084] After the above adjustments, the average continuous 72-hour production capacity of this embodiment is 10.3 kg / hr of nano-calcium carbonate (design value 103%), with a median particle size of 70 nm and σ g =1.35, electron microscopy shows a cubic-spindle mixed morphology, and a specific surface area of ​​27.67 m². 2 / g. After deducting all kinds of consumption, the net equivalent of 0.14t CO2 per ton of product is sealed, achieving "negative carbon" output.

[0085] Figure 3 This is a comparison image of the morphology of samples under different working conditions in the continuous test of this embodiment after drying, under a transmission electron microscope.

[0086] Example 4 This example demonstrates a case of achieving stable production of low-concentration CO2 using calcium carbide slag leachate.

[0087] This embodiment aims to treat low-concentration CO2 sources (~8 vol%) such as cement kiln exhaust gas. The mineralization raw material uses ammonia synthesis process containing NH4. + The high-Ca leachate obtained from leaching calcium carbide slag solid waste combines waste co-processing and carbon reduction benefits. The core challenge under low CO2 partial pressure is the weak mass transfer driving force and the extremely high requirements for reaction selectivity control. This embodiment achieves precise "navigation" of the process through deep coupling of thermodynamic and kinetic models.

[0088] In this embodiment, during system operation, HCO3 is first calculated in real time using online ion data. - With CO3 2- The concentration ratio is a key indicator for determining the selectivity of the mineralization reaction pathway. When the model predicts that this ratio will continue to rise due to temperature fluctuations and exceed 5.0, the system determines that there is excess HCO3.- The tendency for accumulation, potentially leading to the production of NH4HCO3 or Ca(HCO3)2 as byproducts, will automatically reduce the NH4 content on the upstream ore-dissolving side. + The amount of feed will pull the reaction back to the thermodynamically optimal path, which is mainly for the production of CaCO3.

[0089] In subsequent operation of the system, it was found that when the calcium ion concentration was 0.7-0.8 mol / L and the total ammonia concentration was 1.0-2.5 mol / L, the product yield decreased after the pH value decreased from 10.78 to 9.95. Therefore, the system first determined the product yield based on the measured concentration and pH data, and then used the acid-base ionization equilibrium constant K of ammonia and carbonic acid. b K a1 and K a2 Calculate NH3 / NH4 + With HCO3 - / CO3 2- Real-time distribution of H2CO3. Calculation results show that CO3... 2- The distribution coefficient is around 70-80%, placing it at a high level where it has an advantage in product generation. The system then retrieves the K of CaCO3. sp Determining the theoretical limit of CO3 2- After determining the concentration and back-calculating the total carbon concentration, the K concentration required to generate CaCO3 in the solution system is determined. sp The total carbon content is severely insufficient, which offsets the CO3 content. 2- The distribution coefficient was advantageous, therefore the CO2 inlet flow rate needed to be increased. The system transmitted this assessment back to the kinetic data system and compared it, finding that at this point, with the feed temperature close to 50°C, the CO2 solubility decreased sharply, resulting in insufficient total carbon in the solution. The system ultimately recommended increasing the CO2 flow rate or lowering the system temperature.

[0090] In this embodiment, the control system enhances CO2 absorption at low concentrations by dynamically setting the axial temperature gradient of the reactor based on the inlet CO2 concentration and flow rate using an embedded thermodynamic model. The feed section temperature is controlled below 35°C to increase the initial dissolution rate of CO2, and then gradually reduced to 25°C along the tube side to drive the dissolution equilibrium towards CO3 formation. 2- Shift in the direction to maximize Ca 2+ Conversion rate (while closely monitoring the dynamic CaCO3 supersaturation data, ammonia distribution trends, and the difference between theoretical and measured values ​​of key components obtained under theoretical charge conservation, based on online ion data, to determine the degree of deviation of the system from chemical equilibrium or to prevent the formation of ammonia-calcium complexes).

[0091] In addition, the edge computing gateway deployed on the reactor side performs real-time filtering and preliminary analysis on the high-frequency particle data of the dynamic light scattering instrument (DLS). Through a crystallization kinetics micro-model with a root mean square error (RMSE) of less than 0.1, the residence time of the pre-reactor is optimized in reverse to ensure that the seed crystals are in the best growth state.

[0092] Studies have shown that this embodiment successfully and stably produced spindle-shaped nano-calcium carbonate with a purity higher than 98 mol% and uniform morphology by utilizing a low-concentration gas source through a series of multi-parameter synergistic optimization strategies. At the same time, it achieved a carbon sequestration efficiency of about 0.4 tons of CO2 / ton of product, providing a reliable technical solution for the large-scale resource utilization of low-concentration CO2.

[0093] Figure 4 This is a SEM image of the product obtained in Embodiment 2 of the present invention.

[0094] Example 5 This example is a case study on the continuous operation optimization of boiler flue gas treatment using high-calcium fly ash leachate.

[0095] This embodiment uses high-calcium fly ash from a coal-fired power plant as raw material and obtains stable Ca through an acid leaching process. 2+ The extract was used as a carbon source, and nano-calcium carbonate was continuously produced in an industrial-scale tubular nano-microreactor.

[0096] The core of the control method in this embodiment lies in the real-time perception and predictive intervention of complex three-phase systems.

[0097] Specifically, this embodiment integrates an optical fiber probe array and an online ion chromatograph. The former monitors the real-time particle size distribution and movement velocity of nanobubbles within the reactor tube at a frequency of several hundred hertz, ensuring that the main flow diameter remains stable at 50-200 nm. μ The latter has an efficient mass transfer range for m; while the former simultaneously tracks Ca. 2+ NH4 + CO3 2- The second-level changes in the concentration of key ions.

[0098] All multi-source heterogeneous data are timestamped and standardized through a unified data platform. The built-in semi-empirical prediction model integrates real-time pH, temperature, and ion concentration data with a small database of equilibrium constants, solubility, and vapor pressure / dilute solutions to dynamically calculate the instantaneous supersaturation of calcium carbonate in solution under different conditions.

[0099] When the model, based on trend prediction and combined with real-time particle size curve change data provided by the online dynamic light scattering instrument (DLS), determines that oversaturation will exceed the critical threshold of 12 within the next 3-10 minutes, the control system will issue an early warning of particle size distribution deterioration and automatically execute differential control commands: The feed rate of the ammonia metering pump is finely adjusted to enhance the pH buffering capacity of the system and suppress the nucleation rate. At the same time, the opening of the shell-side gas distribution valve is dynamically adjusted based on the CO2 concentration data fed back by the infrared detector at the reactor outlet to optimize the gas-liquid contact efficiency.

[0100] Studies have shown that, through a series of data mining-based feedforward-feedback composite control methods, this embodiment successfully stabilized the CO2 absorption rate at over 85% during continuous operation for hundreds of hours, producing particles with an average diameter of 80 nm and a concentrated particle size distribution (D). 90 / D 10 <2.5 or D 50 High-quality nano-calcium carbonate with / D[4,3]≈1) has achieved a stable transformation from low-value solid waste and waste gas to high-value products.

[0101] Example 6 This embodiment is a case study on the anti-disturbance control of CO2-containing nano-calcium carbonate in blast furnace tail gas mineralized by steel slag leaching solution.

[0102] This embodiment targets the steel industry, specifically the mineralization of blast furnace tail gas (CO2 concentration 20-25 vol%) with significant fluctuations in steel slag leaching solution. The main challenge of this process is the frequent fluctuations in inlet pressure and flow rate, which can easily lead to bubble coalescence and short-circuiting within the nano-microreactor, deteriorating mass transfer and consequently affecting product consistency. The control system in this embodiment addresses this challenge through deep data fusion.

[0103] Specifically, this embodiment integrates cross-sectional imaging data from a small industrial CT scanner with online particle size data from a dynamic light scattering instrument (DLS) to construct a real-time digital twin of the gas-liquid-solid three-phase distribution within the reactor.

[0104] During the test, a blast furnace switching operation caused the tail gas pressure to rise instantly. The monitoring system found that the average diameter of the bubbles increased by more than 30% within 30 seconds.

[0105] The edge computing and data processing module responds promptly: Based on the bubble size and gas holding rate data, it was calculated that the gas-liquid mass transfer area decreased by about 25%, and it was predicted that the macroscopic reaction rate would immediately decrease accordingly.

[0106] The control system then activated the multi-variable coordinated control plan: ①Increase the frequency of the liquid phase circulation pump proportionally to increase the fluid shear force in the tube and break up large bubbles physically. ② The model integrates the supersaturation history curves calculated from the key component concentration data provided by the online ion chromatography and electrochemical system to determine the risk of explosive heterogeneous nucleation. A small amount of pre-prepared crystal form control agent is precisely injected into the reactor to effectively inhibit particle aggregation.

[0107] This embodiment successfully mitigated fluctuations in feed gas concentration and flow rate up to 20 vol% through proactive intervention based on real-time diagnostics, resulting in a stable product specific surface area of ​​25 ± 2 m². 2 Within the range of / g, the robustness of the process under harsh industrial environments is further demonstrated.

[0108] In summary, this invention utilizes industrial solid waste as raw material to achieve the sealing and high-value utilization of CO2 in industrial waste gas. While eliminating environmental pollution, it reduces the carbon emission pressure on enterprises and stably achieves continuous production of nano-calcium carbonate, thereby reducing the production cost of nano-calcium carbonate.

[0109] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for optimizing and controlling process parameters in the continuous production of carbon-negative nano-calcium carbonate in an ammonia-calcium system, characterized in that, The process parameter optimization and control method is applied to the preparation of nano-calcium carbonate using a calcium ion-containing leachate and a carbon dioxide-containing gas in a continuously operating nano-microreactor, and includes the following steps: S1. Real-time acquisition of multi-source heterogeneous data of the gas-liquid-solid three-phase system within a continuously operating nano-microreactor, wherein the data includes at least pH value, Ca... 2+ Concentration, NH3 concentration, NH4 + Concentration, CO3 2- Concentration, HCO3 - Concentration, temperature, bubble size distribution, and gas holdup; S2. After time synchronization and preprocessing of the multi-source heterogeneous data collected in step S1, the data is input into a semi-empirical prediction model. The semi-empirical prediction model is a machine learning model, and a physical consistency constraint term based on the chemical equilibrium constant of the ammonia-calcium system is introduced into the loss function used for its training. The semi-empirical prediction model is configured to output at least the real-time supersaturation of calcium carbonate and the reaction path selectivity evaluation parameters in the reaction system. S3. Based on the real-time supersaturation of calcium carbonate and the reaction path selectivity evaluation parameters of the reaction system output by the semi-empirical prediction model in step S2, process parameter control is performed.

2. The method for optimizing and controlling process parameters for the continuous production of negative carbon nano-calcium carbonate in an ammonia-calcium system according to claim 1, characterized in that, In step S1, the data also includes particle velocity distribution, particle size distribution, bubble velocity distribution, CO2 concentration and flow rate at reactor inlet and outlet, as well as reaction system pressure, liquid phase and gas phase feed flow rate.

3. The method for optimizing and controlling process parameters for the continuous production of carbon-negative nano-calcium carbonate in an ammonia-calcium system according to claim 1, characterized in that, The semi-empirical prediction model is a neural network model.

4. The method for optimizing and controlling process parameters for the continuous production of carbon-negative nano-calcium carbonate in an ammonia-calcium system according to claim 1, characterized in that, The semi-empirical prediction model includes embedded computational units for calculating intermediate variables based on chemical equilibrium equations and chemical kinetics.

5. The method for optimizing and controlling process parameters for the continuous production of negative carbon nano-calcium carbonate in an ammonia-calcium system according to claim 1, characterized in that, In step S2, the chemical equilibrium constant includes the ionization constant of water, K. w The ionization constant K of ammonia b The first-order ionization constant K of carbonic acid a1 With the second-order ionization constant K a2 The solubility product constant of calcium carbonate, K sp .

6. The method for optimizing and controlling process parameters for the continuous production of carbon-negative nano-calcium carbonate in an ammonia-calcium system according to claim 1, characterized in that, In step S2, the reaction pathway selectivity evaluation parameter is the calculated HCO3. - Concentration and CO3 2- The ratio of concentrations; When the ratio exceeds a preset threshold, the NH4 content is adjusted. + Adjust the feed amount and / or lower the system temperature to guide the reaction toward the formation of calcium carbonate as the primary component.

7. The method for optimizing and controlling process parameters for the continuous production of negative carbon nano-calcium carbonate in an ammonia-calcium system according to claim 1, characterized in that, In step S3, when adjusting the process parameters, at least one of the following operations is performed: (a) Adjust the amount of ammonia or ammonium salt added to the reaction system, and / or adjust the inlet pressure or flow rate of carbon dioxide gas; (b) Adjust the temperature of the reaction system or set a temperature gradient along the reactor flow channel; (c) Adjust the circulation rate of the liquid phase in the reactor or inject a crystal form control agent; Among them, when the real-time oversaturation exceeds the preset range, the control operation (a) is executed; When the predicted product particle size deviates from the target range, an adjustment operation (c) is performed.

8. The method for optimizing and controlling process parameters for the continuous production of carbon-negative nano-calcium carbonate in an ammonia-calcium system according to claim 1, characterized in that, The calcium-containing ion-containing leachate is obtained by leaching high-calcium industrial solid waste, which includes any one of high-calcium red mud, carbide slag, high-calcium fly ash, steel slag, and waste concrete. The carbon dioxide-containing gas is industrial waste gas, including any one of boiler flue gas, gasifier tail gas, blast furnace tail gas, and cement kiln tail gas.

9. A process parameter optimization and control system for the continuous production of carbon-negative nano-calcium carbonate in an ammonia-calcium system, characterized in that, include: The data acquisition module is used to collect multi-source heterogeneous data of the gas-liquid-solid three phases in a continuously operating nano-micro reactor in real time. An edge computing and data processing module is used to receive and process data from the data acquisition module and run the semi-empirical prediction model as described in claim 1. The optimization decision and control module is used to receive the predicted values ​​output by the edge computing and data processing module, generate control instructions, and send the control instructions to the execution mechanism; The data acquisition module includes at least an online analyzer for monitoring ion concentration, a probe or imaging device for monitoring bubble parameters, and a gas analyzer for monitoring CO2 concentration and flow rate. The actuator includes at least an ammonia or ammonium salt addition device, a gas flow or pressure regulating device, and a temperature regulating device.

10. A negative carbon nano-calcium carbonate product, characterized in that, The method for optimizing and controlling process parameters for continuous production of negative carbon nano-calcium carbonate in the ammonia-calcium system according to any one of claims 1-8 was used to prepare the product.

Images