Electrolytic chlorine production whole-process intelligent regulation and control system based on digital twinning

Abstract

The invention relates to the technical field of chemical production control, and discloses a digital twinning-based electrolytic chlorine production full-flow intelligent regulation and control system, which comprises a data processing module, a digital twinning module, a digital twinning module and a control module, the digital twinning module is used for constructing a digital twinning model and calculating to obtain the current efficiency, the actual product concentration, the predicted value of the outlet temperature of the electrolytic cell, the electrode state and the salinity; the state estimation module is used for constructing and updating a state vector; the prediction control module is used for constructing a target function and solving the target function to obtain a control instruction; the safety monitoring module is used for executing safety protection measures; by constructing the digital twinborn model and combining a multivariable predictive control algorithm, accurate and stable control over the sodium hypochlorite yield and concentration under the fluctuation condition of multiple parameters such as seawater salinity, temperature and pH is achieved, the control precision is improved, the fluctuation range of the product concentration is narrowed, and the problem that the product quality is affected by seawater quality changes is effectively solved.

Description

Technical Field

[0001] This invention relates to the field of chemical production control technology, and more specifically, to an intelligent control system for the entire electrolytic chlorine production process based on digital twins. Background Technology

[0002] Seawater electrolysis for chlorination is a key process in water treatment and chemical production. It generates sodium hypochlorite by electrolyzing chloride-containing seawater solutions. The stability and efficiency of this process directly impact product quality and industrial energy consumption. While traditional PLC control systems can achieve basic automation, they primarily rely on fixed concentration models and single-parameter linear adjustment, which presents significant limitations under complex operating conditions.

[0003] Chinese Patent CN117742278B discloses an intelligent monitoring and management method and system for sodium hypochlorite production processes. The method includes: obtaining the final sodium hypochlorite concentration, which is manually set according to production requirements; obtaining the initial sodium hypochlorite concentration P1 obtained through an electrolysis generator and the volume L1 of the mixed liquid entering the intermediate reaction tank after electrolysis within time period T; and establishing an intermediate sodium hypochlorite concentration model, where the intermediate sodium hypochlorite concentration is the sodium hypochlorite concentration in the mixed liquid formed after the reaction of sodium hydroxide solution and chlorine gas. This invention adds a process for converting the generated chlorine gas into sodium hypochlorite, and establishes an intermediate sodium hypochlorite concentration model based on this process structure. This enables control of the concentration of sodium hydroxide participating in the reaction, ultimately achieving full utilization of the chlorine gas generated by electrolysis and saving resources.

[0004] However, key parameters of seawater such as salinity, temperature, and pH can fluctuate significantly due to various factors. High summer temperatures can raise seawater temperatures, affecting electrolysis efficiency and product stability. Tidal changes can cause fluctuations in seawater salinity, directly impacting ion concentration during electrolysis. Current technologies primarily rely on fixed concentration models and simple linear fitting methods for control, which cannot adapt to these complex and changing seawater quality conditions in real time. When seawater composition changes, the system cannot adjust electrolysis parameters promptly, leading to decreased electrolysis efficiency, increased energy consumption, unstable sodium chlorate production, and deviations from optimal equipment operating conditions, ultimately affecting equipment lifespan. Summary of the Invention

[0005] The purpose of this invention is to solve the above-mentioned problems. Therefore, this invention provides an intelligent control system for the entire process of electrolytic chlorine production based on digital twins.

[0006] This invention provides an intelligent control system for the entire electrolytic chlorine production process based on digital twins, comprising:

[0007] The data processing module is used to collect and preprocess process parameters for the entire chemical production process in real time.

[0008] The digital twin module constructs a digital twin model based on the process parameters. The digital twin model calculates the current efficiency, actual product concentration, predicted value of electrolytic cell outlet temperature, electrode state, and salinity.

[0009] The state estimation module constructs and updates the state vector based on the process parameters, current efficiency, actual product concentration, electrode state, and salinity.

[0010] The predictive control module constructs an objective function based on the digital twin model and the state vector, and solves the objective function to obtain control commands;

[0011] The safety monitoring module performs safety verification on control commands and executes safety protection measures based on the predicted value of the electrolytic cell outlet temperature and the state vector.

[0012] Furthermore, the digital twin model includes an electrochemical model, a thermodynamic model, an electrode state model, and a salinity estimation model, among which:

[0013] The actual product concentration is calculated by the electrochemical model. The actual product concentration is the product of the current efficiency and the electrolysis current, and then divided by the product of the preset electron transfer number, the preset Faraday constant and the seawater flow rate. The current efficiency is calculated by the efficiency calculation model.

[0014] The thermodynamic model calculates the predicted value of the electrolyzer outlet temperature using the energy balance equation;

[0015] The electrode state is calculated by the electrode state model. The electrode state at the next moment is the product of the preset state attenuation coefficient and the current electrode state, plus the product of the preset scale growth coefficient and the voltage term. The voltage term is the maximum value between 0 and the voltage difference. The voltage difference is the difference between the current tank voltage and the reference voltage.

[0016] The salinity estimation model uses a multiple linear regression method to calculate the salinity of seawater. The salinity is calculated by adding a preset second constant term to the product of the conductivity coefficient and conductivity, the preset temperature coefficient and temperature, and the preset interaction term coefficient of conductivity and temperature to the product of conductivity and temperature.

[0017] Further, the reference voltage is the sum of the following: a preset first constant term, the product of the first-order coefficient of the preset salinity and salinity, the product of the first-order coefficient of the preset temperature and temperature, the product of the first-order coefficient of the preset current and current, the product of the quadratic coefficient of the preset salinity and the square of salinity, the product of the quadratic coefficient of the preset temperature and the square of temperature, the product of the quadratic coefficient of the preset current and the square of current, the product of the interaction coefficient of the preset salinity and temperature and the product of salinity and temperature, the product of the interaction coefficient of the preset salinity and current and the product of the interaction coefficient of the preset temperature and current and the product of temperature and current.

[0018] Furthermore, the energy balance equation is: the product of seawater density, seawater specific heat capacity, and seawater flow rate, multiplied by the difference between the predicted value of the electrolytic cell outlet temperature and the electrolytic cell inlet temperature, equals the electrolysis power minus the heat dissipation loss power; the electrolysis power is the product of cell voltage and electrolysis current, and the heat dissipation loss power is the product of the preset total heat transfer coefficient and the preset heat exchange area, multiplied by the difference between the average temperature and the ambient temperature, and the average temperature is the arithmetic mean of the predicted values ​​of the electrolytic cell inlet temperature and the electrolytic cell outlet temperature.

[0019] Furthermore, the control commands received include:

[0020] The objective function is to minimize the sum of performance indicators at each time point in the prediction time domain. The performance indicators include the product of the preset concentration tracking weight coefficient and the square of the difference, the product of the preset energy consumption optimization weight coefficient and the predicted power, the product of the preset electrode protection weight coefficient and the electrode state, and the product of the preset control stability weight coefficient and the square of the L2 norm of the control variable increment vector. Here, the square of the difference is equal to the square of the difference between the actual product concentration and the preset target concentration setting value, the predicted power is equal to the product of the cell voltage and the electrolysis current setting value, and the control variable increment vector is the change of the control variable between adjacent time points.

[0021] A multivariable predictive control algorithm is used to solve the objective function under constraints to obtain control commands.

[0022] Furthermore, the control command includes multiple control variables, including the electrolysis current setpoint, seawater flow rate setpoint, seawater booster pump frequency setpoint, and acid mist absorption fan speed setpoint.

[0023] Furthermore, the constraints include safety constraints, process constraints, and control constraints; among which, safety constraints include electrolyzer outlet temperature constraints, hydrogen concentration constraints, flow rate lower limit constraints, and voltage upper limit constraints; process constraints include sodium hypochlorite storage tank level constraints, electrolysis current range constraints, and seawater booster pump frequency range constraints; and control constraints include control variable increment limits and control variable range constraints.

[0024] Furthermore, the safety protection measures include a hydrogen concentration safety control strategy and a temperature safety control strategy; wherein, the hydrogen concentration safety control strategy achieves graded control of hydrogen concentration through fan speed adjustment and electrolysis current control, and the hydrogen concentration control target is set to be less than a preset normal hydrogen threshold under normal operating conditions; the temperature safety control strategy is based on a thermodynamic model, and the electrolyzer outlet temperature control target is set to be less than a preset safety threshold under normal operating conditions.

[0025] Furthermore, the state estimation module uses an extended Kalman filter algorithm to update the state vector. The state vector includes state variables, such as salinity, current efficiency, electrode state, and actual product concentration.

[0026] Furthermore, the process parameters include flow rate parameters, electrical parameters, temperature parameters, pressure parameters, liquid level parameters, water quality parameters, and safety monitoring parameters; wherein, the flow rate parameters include seawater flow rate and sodium hypochlorite product delivery flow rate, the electrical parameters include electrolysis current and cell voltage, the temperature parameters include observed values ​​of the electrolytic cell inlet temperature and electrolytic cell outlet temperature, the pressure parameters include the electrolytic cell inlet pressure, the liquid level parameters include the liquid level of the sodium hypochlorite storage tank, the liquid level of the pickling tank, and the liquid level of the concentrated acid tank, the water quality parameters include conductivity and pH value, and the safety monitoring parameters include hydrogen concentration, equipment vibration signal, and electrical fault signal.

[0027] The beneficial effects of this invention are as follows: By constructing a digital twin model and combining it with a multivariate predictive control algorithm, this invention achieves precise and stable control of sodium hypochlorite production and concentration under fluctuating seawater salinity, temperature, pH, and other parameters. Compared with the traditional single-variable PID control method, it improves control accuracy, reduces the product concentration fluctuation range, and effectively solves the problem of the impact of seawater quality changes on product quality. Through a deep learning-based current efficiency prediction model, it can accurately predict the electrolysis efficiency under different operating conditions. The multivariate predictive control algorithm optimizes key parameters such as electrolysis current and seawater flow rate in real time, minimizing energy consumption while ensuring product quality. In practical applications, this invention reduces system energy consumption compared to traditional control methods, effectively saving annual electricity costs.

[0028] This invention monitors electrode scaling and aging in real time using an electrode state model, providing early warnings of equipment maintenance needs and preventing sudden failures. A dual-layer control coordination mechanism ensures automatic switching to a safe mode when intelligent control fails, improving system reliability. By mapping the physical equipment state in real time using a digital twin model and continuously optimizing model parameters using online learning algorithms, intelligent diagnosis of equipment status and fault prediction are achieved. The electrode cleaning cycle can be intelligently adjusted according to actual conditions, reducing maintenance costs and effectively improving equipment utilization compared to fixed-cycle maintenance.

[0029] This invention solves the technical challenge of directly measuring seawater salinity by employing an extended Kalman filter algorithm to accurately estimate key state variables such as salinity and current efficiency. The online learning module adaptively adjusts model parameters based on changing operating conditions, ensuring good control performance under different seasons and sea conditions. A complete data acquisition, processing, and analysis system is established to collect process parameters in real time, including flow rate, electrical parameters, temperature, pressure, liquid level, water quality, and safety parameters, achieving industry-leading sampling accuracy. Data mining and trend analysis provide a scientific basis for production optimization and management decisions, enhancing the enterprise's digital management level. Precise control and energy consumption optimization reduce energy consumption and waste generation in the electrolysis process. Intelligent acid mist treatment and hydrogen monitoring ensure environmental safety, complying with national environmental protection policies. Digital management reduces manual operation and paper records, reflecting the development concept of green and intelligent manufacturing.

[0030] This invention achieves effective improvements in multiple aspects, such as control precision, energy consumption optimization, safety and reliability, and intelligent operation and maintenance, through the deep integration of digital twin technology and intelligent control algorithms. It provides a complete technical solution for the intelligent upgrading of the electrolytic chlorine production industry and has significant technical value and broad application prospects. Attached Figure Description

[0031] Figure 1 This is a module example diagram of the intelligent control system for the entire electrolytic chlorine production process based on digital twins according to the present invention;

[0032] Figure 2 This is an example diagram of the digital twin model for constructing the intelligent control system for the entire electrolytic chlorine production process based on digital twins according to the present invention;

[0033] Figure 3 This is an example diagram of the construction state vector of the intelligent control system for the entire electrolytic chlorine production process based on digital twins according to the present invention;

[0034] Figure 4 This is an example diagram illustrating the control commands obtained by the intelligent control system for the entire electrolytic chlorine production process based on digital twins according to the present invention. Detailed Implementation

[0035] The subject matter described herein will now be discussed with reference to exemplary embodiments. It should be understood that these embodiments are discussed only to enable those skilled in the art to better understand and implement the subject matter described herein, and changes may be made to the function and arrangement of the elements discussed without departing from the scope of this specification. Various processes or components may be omitted, substituted, or added as needed in the examples. Furthermore, features described in some examples may be combined in other examples.

[0036] A digital twin-based intelligent control system for the entire electrolytic chlorine production process, such as Figure 1 As shown, it includes:

[0037] The data processing module 101 is used to collect and preprocess the process parameters of the entire chemical production process in real time; the chemical production refers to the electrolytic production of chlorine.

[0038] As the foundational layer of the entire intelligent control system, the data processing module is responsible for real-time acquisition of process parameter information for the entire electrolytic chlorine production process, providing high-quality data support for subsequent digital twin modeling, state estimation, and predictive control.

[0039] The data processing module uses a PLC as the basic control platform and achieves real-time acquisition of field signals through distributed I / O modules. The real-time process data acquired by the system covers key parameters of the entire electrolytic chlorine production process, ensuring comprehensive monitoring of the production process. Process data includes flow parameters, electrical parameters, temperature parameters, pressure parameters, liquid level parameters, water quality parameters, and safety monitoring parameters.

[0040] Regarding flow parameters, the system collects flow signals such as seawater flow and sodium hypochlorite product delivery flow in real time, using electromagnetic flowmeters or turbine flowmeters for measurement. The sampling frequency is preset to 1Hz, and the measurement accuracy reaches ±0.5%, providing accurate feedback information for flow control. Seawater flow refers to the flow rate of raw seawater entering the electrolytic cell, and sodium hypochlorite product delivery flow refers to the flow rate of sodium hypochlorite product supplied to downstream users via a continuous impact dosing pump. Electrical parameter acquisition includes key electrical signals such as electrolysis current and cell voltage, which are measured in real time using current and voltage transmitters at a sampling frequency of 10Hz. The current measurement accuracy is ±0.2%, and the voltage measurement accuracy is ±0.1%, ensuring precise control of the electrolysis process.

[0041] Temperature parameter monitoring includes observed temperatures such as the electrolyzer inlet temperature and electrolyzer outlet temperature. Real-time measurements are performed using platinum resistance temperature sensors with a sampling frequency of 0.5 Hz and a measurement accuracy of ±0.1°C, providing accurate temperature data for the thermodynamic model. The electrolyzer inlet temperature refers to the temperature of the seawater before it enters the electrolyzer, while the observed electrolyzer outlet temperature refers to the temperature of the seawater leaving the electrolyzer after the electrolysis reaction, as measured in real-time by the sensor. Pressure parameters include pressure signals such as the electrolyzer inlet pressure, measured in real-time by a pressure transmitter with a sampling frequency of 1 Hz and a measurement accuracy of ±0.25%, ensuring stable system pressure control. The electrolyzer inlet pressure refers to the pressure of the seawater before it enters the electrolyzer and is used to monitor the operating status of the seawater supply system.

[0042] The system collects level parameters, including levels in the sodium hypochlorite storage tank, pickling tank, and concentrated acid tank. These levels are measured in real-time using ultrasonic or radar level gauges at a sampling frequency of 0.2 Hz and a measurement accuracy of ±1 mm, providing reliable data for level control and safety monitoring. Regarding water quality parameters, the system monitors conductivity and pH values ​​in real-time using online conductivity meters and pH meters. The sampling frequency is 0.1 Hz, with conductivity measurement accuracy of ±1% and pH measurement accuracy of ±0.1 pH, providing fundamental data for water quality assessment and salinity estimation.

[0043] Safety monitoring parameters include hydrogen concentration, equipment vibration signals, and electrical fault signals. These are monitored in real time using hydrogen detectors, vibration sensors, and fault detection devices, with a sampling frequency of 2Hz, to ensure the safe and reliable operation of the production process and to promptly detect and address potential safety hazards.

[0044] The data processing module covers the main equipment and corresponding data acquisition points for the entire electrolytic chlorine production process, forming a complete monitoring network. The seawater booster pump, as the system's power source, is responsible for providing a stable seawater flow rate. Precise regulation of the seawater flow rate is achieved through frequency conversion control. It is equipped with flow sensors, pressure sensors, current sensors, and vibration sensors to comprehensively monitor the pump's operating status. The automatic backwash filter plays a crucial role in removing suspended solids and impurities from the seawater. It is equipped with differential pressure sensors, backwash status sensors, and fault alarm sensors to ensure filtration effectiveness and normal equipment operation.

[0045] The sodium hypochlorite generator, as the core equipment for electrolytic chlorine production, includes an electrolytic cell and electrode assembly. It is equipped with temperature sensors, pressure sensors, liquid level sensors, and hydrogen concentration sensors to achieve real-time monitoring of key parameters in the electrolysis process. The rectified power supply provides the necessary DC power for the electrolysis process. Both the output electrolysis current and cell voltage are adjustable. It is equipped with current transmitters, voltage transmitters, a power meter, and a fault detection device to ensure the stability and reliability of the power output.

[0046] Sodium hypochlorite storage tanks are used to store the generated sodium hypochlorite solution. They are equipped with level sensors, temperature sensors, and concentration monitoring devices to monitor product quality and storage status. Continuous / impact dosing pumps control the outgoing flow rate of sodium hypochlorite, and are equipped with flow sensors, pressure sensors, and operating status sensors to ensure precise control of the product delivery process.

[0047] The pickling system includes a pickling tank and a concentrated acid tank, specifically designed for electrode cleaning and maintenance. It is equipped with level sensors, temperature sensors, pH sensors, and acid mist concentration sensors to ensure the safety and effectiveness of the cleaning process. The acid mist absorber and fan system are responsible for handling the acid mist generated during the pickling process. It is equipped with wind speed sensors, differential pressure sensors, and acid mist concentration sensors to ensure environmental safety and operator health.

[0048] The data preprocessing function covers key aspects such as signal filtering, data verification, data caching, and data compression to ensure data quality and system efficiency. For signal filtering, the system uses digital filters to remove high-frequency noise and improve signal quality. For rapidly changing electrical signals, a low-pass filter with a cutoff frequency of 5Hz is used, while for slowly changing temperature and liquid level signals, a moving average filter with a window length of 10 sampling points is used to effectively suppress noise interference.

[0049] Data validation identifies anomalous data through reasonableness checks, ensuring data reliability. The system establishes normal range thresholds for each parameter; data exceeding these ranges is marked as anomalous. Continuous checks identify data jumps and sensor malfunctions, and redundant verification is implemented. Backup sensors are configured for critical parameters for cross-validation, improving data accuracy and reliability.

[0050] The data caching function establishes a historical data buffer, supporting trend analysis and historical queries. The system uses a circular buffer to store historical data for the most recent 24 hours, with a sampling interval of 1 minute. An event trigger buffer is also established to record detailed data for 30 minutes before and after alarms and abnormal events, providing data support for fault analysis and system optimization. Data compression employs a rate-of-change compression algorithm to reduce storage space and improve system efficiency. For parameters that change slowly, new data points are recorded only when the value change exceeds a preset threshold. For parameters that change periodically, a differential compression algorithm is used to reduce data redundancy.

[0051] In terms of communication interfaces, the system is equipped with industrial Ethernet for high-speed data exchange with the edge computing layer, achieving a data transmission rate of 100Mbps. The communication protocol uses TCP / IP to ensure real-time and reliable data transmission. The system uses the Modbus RTU / TCP protocol to communicate with third-party devices, supporting up to 32 slave devices and providing excellent device compatibility. The system provides an OPC UA server to offer a standardized data interface to the host computer system, supporting real-time data access and historical data querying to meet different levels of data needs.

[0052] The edge computing layer consists of industrial computers or embedded controllers, interconnected with the PLC system via the industrial Ethernet protocol. This layer runs digital twin models, state estimators, and multivariable predictive control algorithms to achieve intelligent optimization control. Edge computing nodes include core functional modules such as data processing, digital twin, state estimation, predictive control, online learning, and safety monitoring modules, forming a complete intelligent control system.

[0053] The real-time process data collected and processed by the data processing module is preprocessed and then transmitted to module 102 via industrial Ethernet for digital twin modeling. At the same time, it provides observation data input for the state estimation algorithm of module 103.

[0054] Digital twin module 102, based on the process parameters, constructs a digital twin model. The digital twin model calculates current efficiency, actual product concentration, predicted electrolyzer outlet temperature, electrode state, and salinity, specifically as follows: Figure 2 As shown.

[0055] Based on the real-time process data provided by Module 101, this module constructs a digital twin model of the electrolytic chlorine production process. Through multiphysics coupling modeling, a comprehensive digital twin system covering electrochemical reactions, heat and mass transfer, electrode evolution, and water quality changes is established, providing an accurate description of the process mechanism for state estimation in Module 103 and predictive control in Module 104. The digital twin model includes an electrochemical model, a thermodynamic model, an electrode state model, and a salinity estimation model.

[0056] The construction of the digital twin model follows a progressive modeling approach, moving from fundamental physical processes to system integration. First, a hierarchical control architecture is established, clarifying the digital twin model's role and operational mechanism within the entire intelligent control system. Based on this, an electrochemical model describing the core reaction process of chlorine production by electrolysis is constructed. This model, grounded in Faraday's law and combined with a deep learning-based efficiency calculation model, accurately predicts the formation rate and concentration of sodium hypochlorite. The heat generated during the electrochemical reaction is described using a thermodynamic model based on the energy balance principle, considering electrolysis power input and heat dissipation losses to achieve accurate prediction of the system's temperature distribution. Electrodes, as the carriers of the electrochemical reaction, directly affect system performance; therefore, an electrode state model is established, using a data-driven approach to describe the scaling and aging evolution of the electrodes. Seawater salinity, a key parameter affecting current efficiency, cannot be directly measured; therefore, an accurate estimation is achieved through an online regression model based on conductivity and temperature.

[0057] In the intelligent control system for electrolytic chlorine production, the digital twin module acts as a bridge between the physical and digital worlds, and its architecture design directly affects the control performance and operational effectiveness of the entire system. This embodiment adopts a hierarchical control architecture, embedding the digital twin model into a three-layer architecture including a field device layer, an edge computing layer, and a digital twin layer, achieving end-to-end integration from data acquisition to intelligent control.

[0058] The field device layer, serving as the physical foundation of the system, is responsible for the actual operation and basic data acquisition of the electrolytic chlorine production equipment. It implements safety interlock protection functions through a PLC system, ensuring safe shutdown under any abnormal conditions. The edge computing layer, as the core of intelligent control, runs digital twin models and intelligent control algorithms. Based on real-time data and model predictions, it makes optimized decisions and exchanges data with the field device layer via high-speed industrial Ethernet. The digital twin layer constructs electrochemical, thermal, and fluid coupling models, providing a virtual mapping of the system state and offering accurate process prediction capabilities for upper-level control algorithms.

[0059] The system employs a two-layer control coordination mechanism to ensure the organic integration of intelligent control and basic control. The upper-layer intelligent control operates at the edge computing layer, executing digital twin models and multivariable predictive control algorithms. Based on a global optimization objective, it calculates the optimal control setpoint and transmits it to the PLC system in real time via industrial Ethernet. The lower-layer basic control operates at the PLC layer, receiving the setpoints from the upper-layer intelligent control, executing traditional closed-loop control strategies, while maintaining all safety interlocks and alarm functions. It automatically switches to a preset backup control mode when the upper-layer intelligent control fails.

[0060] The two-layer control coordination mechanism, through carefully designed interface protocols and state synchronization mechanisms, ensures a smooth transition of setpoints and avoids adverse effects on the system from control shocks. State information is uploaded to the edge computing layer in real time, maintaining synchronization between the digital twin model and the physical system. Automatic degradation processing is implemented when anomalies are detected, ensuring the safe and reliable operation of the system.

[0061] The core of the electrolytic chlorine production process lies in the electrochemical reaction that converts chloride ions in seawater into sodium hypochlorite. The reaction efficiency directly determines the system's production capacity and energy consumption. The electrochemical model, a core component of the digital twin system, is built upon Faraday's law of electrolysis and employs a deep learning-based efficiency calculation model to provide the reaction kinetics foundation for the entire digital twin model.

[0062] The theoretical basis of the electrolysis reaction follows Faraday's law, where the molar rate of chlorine formation is directly proportional to the electrolysis current. The molar rate of chlorine formation equals the electrolysis current divided by the product of a preset electron transfer number and a preset Faraday constant. Here, the electrolysis current is in amperes; the preset electron transfer number is assumed to be 2, determined based on the electrochemical reaction mechanism of chloride ions losing two electrons to form chlorine; and the preset Faraday constant is assumed to be 96485 coulombs per mole, determined based on fundamental physicochemical constants. This theoretical relationship provides a basic computational framework for the electrochemical model. However, the actual electrolysis process is affected by various factors, necessitating the introduction of the concept of current efficiency for correction.

[0063] The calculation of actual product concentration needs to consider the influence of current efficiency. The actual product concentration is equal to the product of current efficiency and electrolysis current, divided by the product of the preset electron transfer number, the preset Faraday constant, and the seawater flow rate, where the seawater flow rate is in cubic meters per second. Current efficiency reflects the degree to which electrical energy is effectively converted into chemical energy during electrolysis and is an important link connecting the electrochemical model with other sub-models.

[0064] Current efficiency is influenced by a combination of factors, including salinity, temperature, current density, and electrode condition. Changes in these factors directly affect the efficiency of the electrolysis reaction. To accurately predict the variation of current efficiency under complex operating conditions, this embodiment employs a deep learning-based efficiency calculation model to achieve precise calculation of current efficiency.

[0065] The efficiency calculation model employs a deep neural network architecture, capturing the complex mapping relationship between input parameters and current efficiency through multi-layer nonlinear transformations. The model's input parameters include key process parameters such as salinity, observed electrolyzer inlet and outlet temperatures, electrolytic current, cell voltage, current density, electrode state, seawater flow rate, conductivity, and pH. These input parameters are standardized through a data preprocessing module to eliminate the influence of different parameter dimensions and numerical ranges, ensuring the stability and convergence of model training.

[0066] The efficiency calculation model's network structure includes an input layer, three hidden layers, and an output layer. The input layer receives standardized input parameters. The first hidden layer comprises sixty-four neurons, employing a modified linear unit activation function (MLU) to extract basic features from the input parameters. The second hidden layer comprises thirty-two neurons, also employing the MLU, to further extract higher-order feature combinations. The third hidden layer comprises sixteen neurons, employing the MLU to achieve deep feature fusion. The output layer comprises one neuron, employing the sigmoid activation function, and outputs a predicted current efficiency value, ranging from 0 to 1.

[0067] The training process of the efficiency calculation model is divided into two stages: offline training and online updating. The offline training stage utilizes historical operating data to construct a training dataset, which includes combinations of input parameters under different operating conditions and their corresponding actual current efficiency measurements. Training data is obtained through periodic product concentration analysis and electrolysis current recording, and the actual current efficiency values ​​are calculated according to Faraday's law. The training dataset is divided into a training set and a validation set in an 8:2 ratio. The training set is used for model parameter optimization, while the validation set is used for model performance evaluation and overfitting detection.

[0068] The model training employed backpropagation and an adaptive moment estimation optimizer, with a learning rate of 0.01, a batch size of 32, and a maximum training epoch of 1000 epochs. The mean squared error function (MSE) was used as the loss function to measure the deviation between the predicted and actual current efficiency. An early stopping strategy was employed during training: training was halted when the validation set loss stopped decreasing for 20 consecutive epochs to prevent overfitting. A learning rate decay strategy was also used: the learning rate was halved when the validation set loss stagnated to improve model convergence accuracy.

[0069] During the online update phase, real-time obtained test data is used to incrementally learn the model, maintaining its adaptability to changes in operating conditions. When new product concentration test results are obtained, the corresponding actual current efficiency is calculated and compared with the model's predicted value to calculate the prediction error. If the prediction error exceeds a preset threshold, the online update mechanism for model parameters is triggered. The online update employs a mini-batch gradient descent algorithm, with the learning rate set to one-tenth of the offline training learning rate to ensure the stability of the update process. During the update process, a sliding window of the most recent 1000 samples is retained to prevent the model from forgetting historical knowledge.

[0070] The predicted output current efficiency of the efficiency calculation model is directly used to calculate the actual product concentration in the electrochemical model, replacing traditional empirical formulas or simplified models. This predicted value is also fed back to the electrode condition model as an important indicator for electrode performance evaluation. A continuous decrease in the predicted current efficiency indicates potential deterioration of the electrode condition, necessitating cleaning or maintenance.

[0071] The output of the electrochemical model not only provides electrolysis power input for the thermodynamic model, but also provides current density information for the electrode state model and reaction rate reference for the salinity estimation model, realizing data flow and information sharing among the sub-models.

[0072] The electrochemical reactions in the electrolytic chlorine production process are accompanied by significant thermal effects. Changes in system temperature not only affect the efficiency of the electrolytic reaction but also the safe operation of the equipment and the stability of product quality. A thermodynamic model, based on the principle of energy balance, is closely coupled with the electrochemical model to jointly describe the energy conversion and temperature distribution patterns during the electrolysis process.

[0073] The thermodynamic model uses the electrolysis power provided by the electrochemical model as the main heat source input, while also considering the heat loss to the environment, to establish a complete energy balance equation. The energy balance relationship of the system is expressed as follows: the enthalpy change of seawater equals the electrolysis power input minus the heat loss power. Specifically, this is expressed as the product of seawater density, seawater specific heat capacity, and seawater flow rate, multiplied by the difference between the predicted outlet temperature and the inlet temperature of the electrolyzer, which equals the electrolysis power minus the heat loss power.

[0074] Electrolysis power is directly derived from the calculation results of the electrochemical model and is equal to the product of cell voltage and electrolysis current, expressed in watts. It reflects the conversion of electrical energy into heat energy during the electrochemical reaction. Heat loss power describes the heat transfer process from the system to the environment and is proportional to the difference between the system temperature and the ambient temperature. It is calculated by multiplying the preset total heat transfer coefficient by the preset heat exchange area, and then multiplying by the difference between the average temperature and the ambient temperature.

[0075] The preset overall heat transfer coefficient is a parameter representing the heat transfer characteristics of the equipment, with a default value of 50 watts per square meter (Kelvin). It is determined through equipment thermal performance testing and reflects the equipment's heat dissipation capacity. The preset heat exchange area is the heat exchange surface area of ​​the equipment, with a default value of 20 square meters, determined based on the equipment's structural design. The average temperature is the arithmetic mean of the predicted inlet and outlet temperatures of the electrolytic cell, and the ambient temperature is the actual temperature of the surrounding environment.

[0076] The thermodynamic model calculates the predicted electrolyzer outlet temperature using the energy balance equation. This predicted value is compared with the actual observed electrolyzer outlet temperature observed by module 101, used for online correction of model parameters and real-time monitoring of system status. The deviation between the predicted and observed outlet temperatures reflects changes in model accuracy and system operating status. When the deviation exceeds a preset threshold, an adaptive update mechanism for model parameters is triggered. Simultaneously, the predicted electrolyzer outlet temperature provides the multivariate predictive control algorithm with trend information on future temperature changes, supporting the formulation of temperature-related safety constraints and control strategies.

[0077] The temperature prediction results of the thermodynamic model directly affect the temperature correction coefficient of the current efficiency in the electrochemical model. At the same time, it provides temperature environment information for the electrode state model and temperature compensation parameters for the salinity estimation model, realizing the coupled description of thermodynamic processes and other physical processes.

[0078] As the direct carrier of the electrolytic reaction, the surface state of the electrode directly affects the current efficiency and system performance. During long-term operation, electrodes undergo scaling and aging, leading to increased resistance and mass transfer resistance, ultimately impacting the electrolytic reaction. The electrode state model employs a data-driven modeling approach, combining output information from electrochemical and thermodynamic models to describe the dynamic evolution of the electrode state.

[0079] The evolution of electrode states is influenced by various factors, including current density, temperature, and seawater quality. The model uses a state evolution equation to describe the temporal changes in electrode states. The electrode state at the next moment is equal to the product of the preset state decay coefficient and the current electrode state, plus the product of the preset scaling growth coefficient and the voltage term. The voltage term is the maximum of 0 and the voltage difference, which is the difference between the current tank voltage and the reference voltage. The preset state decay coefficient reflects the natural aging process of the electrode and is a value between 0 and 1, with a default value of 0.995, determined through fitting electrode aging experimental data. The preset scaling growth coefficient reflects the growth rate of scaling on the electrode surface and is a value greater than 0, with a default value of 0.001, determined through fitting electrode scaling experimental data.

[0080] The reference voltage function describes the voltage characteristics of the electrode under ideal conditions. It is the theoretical voltage value under given salinity, temperature, and current conditions, reflecting the correlation between the electrode state model and other sub-models. The reference voltage function is established using a multiple regression method, comprehensively considering the influence of factors such as salinity, temperature, and current. The reference voltage is equal to the sum of the following: the product of the coefficient of the first term of the preset salinity with salinity; the product of the coefficient of the first term of the preset temperature with temperature; the product of the coefficient of the first term of the preset current with current; the product of the coefficient of the quadratic term of the preset salinity with the square of salinity; the product of the coefficient of the quadratic term of the preset temperature with the square of temperature; the product of the coefficient of the quadratic term of the preset current with the square of current; the product of the coefficient of the interaction term of preset salinity and temperature with salinity and temperature; the product of the coefficient of the interaction term of preset salinity and current with salinity and current; and the product of the coefficient of the interaction term of preset temperature and current with temperature and current.

[0081] The preset regression coefficients include the preset first constant term, the preset first-order coefficient of salinity, the preset first-order coefficient of temperature, the preset first-order coefficient of current, the preset second-order coefficient of salinity, the preset second-order coefficient of temperature, the preset second-order coefficient of current, the preset interaction coefficients of salinity and temperature, the preset interaction coefficients of salinity and current, and the preset interaction coefficients of temperature and current, which are determined by fitting historical operating data.

[0082] The output of the electrode state model is directly fed back to the electrochemical model as an electrode state correction coefficient in the current efficiency calculation, forming a closed-loop feedback relationship between electrode state and current efficiency. Simultaneously, electrode state information provides crucial information for system maintenance decisions; when the electrode state deteriorates to a certain extent, the system can automatically trigger cleaning or replacement procedures.

[0083] Seawater salinity is a key parameter affecting the efficiency of chlorine electrolysis. However, due to the high cost, complex maintenance, and susceptibility to contamination and corrosion in industrial environments, online salinity sensors cannot be directly measured online. A salinity estimation model, based on easily measurable and reliable conductivity and temperature parameters, achieves accurate estimation of seawater salinity through online regression methods, providing crucial input parameters for the electrochemical model. Conductivity sensors offer advantages such as fast response, high measurement accuracy, and low maintenance costs, and can operate stably for extended periods in harsh industrial environments, making them an ideal choice for indirect salinity measurement.

[0084] The seawater salinity estimation model employs a multiple linear regression method to establish a quantitative relationship between conductivity, temperature, and seawater salinity. Seawater conductivity and salinity are strongly correlated, but are also significantly affected by temperature; therefore, a regression model including conductivity, temperature, and their interaction term is required. Seawater salinity is calculated by adding a preset second constant term to the product of a preset conductivity coefficient and the conductivity value, a preset temperature coefficient and the temperature value, and a preset interaction term coefficient between conductivity and temperature to the product of conductivity and temperature.

[0085] The model parameters include four regression coefficients: a preset second constant, a preset conductivity coefficient, a preset temperature coefficient, and a preset interaction coefficient between conductivity and temperature. Online adaptive updates are achieved using recursive least squares. The reference salinity value required for parameter updates is obtained through periodic offline laboratory analysis. Specifically, seawater samples are collected daily and sent to the laboratory for salinity measurement using a standard salinity meter or refractometer, achieving a measurement accuracy of one-thousandth of a salinity unit. While this offline method cannot provide real-time salinity data, it provides an accurate reference benchmark for the online estimation model, used for periodic calibration and validation of the model parameters. The parameter update process consists of three steps. First, the Kalman gain vector is calculated by multiplying the product of the covariance matrix and the regression vector by the unit value, multiplying the transpose of the regression vector by the covariance matrix, and then multiplying by the regression vector again. Then, the parameter estimation vector is updated by adding the current parameter estimation vector to the product of the Kalman gain vector, the reference salinity value, and the transpose of the regression vector multiplied by the current parameter estimation vector. Finally, update the covariance matrix. The new covariance matrix is ​​obtained by multiplying the inverse of the preset forgetting factor by the current covariance matrix, subtracting the Kalman gain vector multiplied by the transpose of the regression vector, and then multiplying by the current covariance matrix.

[0086] The parameter estimation vector includes four regression coefficients. The regression vector includes a preset second constant term, conductivity value, temperature value, and the product of conductivity and temperature. The preset forgetting factor is a parameter that controls the weight of historical data. The default value is 0.98, which is determined through online learning performance optimization. A smaller forgetting factor allows the model to adapt to parameter changes more quickly, but may increase estimation noise.

[0087] The output of the salinity estimation model is directly input into the electrochemical model as an important parameter for current efficiency calculation. At the same time, it provides salinity information for the reference voltage calculation in the electrode state model, realizing the organic combination of water quality parameters and electrolysis process.

[0088] The digital twin model, through the organic integration of the four sub-models mentioned above, forms a comprehensive digital twin system describing the entire process of chlorine production by electrolysis. The electrochemical model provides the basis for reaction kinetics, the thermodynamic model describes the energy conversion process, the electrode state model reflects the evolution of equipment state, and the salinity estimation model enables online estimation of key parameters. The sub-models are tightly coupled through parameter transfer and information sharing, jointly providing the state transition function and observation function for the extended Kalman filter in module 103, and providing the process prediction model for the multivariate predictive control algorithm in module 104. The model parameters are adaptively updated through the online learning mechanism in module 106, ensuring that the digital twin model always maintains a high degree of consistency with the actual physical system.

[0089] The state estimation module 103 constructs and updates the state vector based on the process parameters, current efficiency, actual product concentration, electrode state, and salinity, as detailed below. Figure 3 As shown.

[0090] Based on the real-time process data provided by module 101 and the digital twin model constructed by module 102, this module uses the extended Kalman filter algorithm to estimate key state variables that cannot be directly measured in the electrolytic chlorine production process in real time, providing accurate internal system state information for the multivariate predictive control of module 104, and ensuring that control decisions are based on a complete understanding of the system state.

[0091] The state estimation module plays a crucial bridging role in the entire intelligent control system, connecting the digital twin model and the multivariate predictive control algorithm. Since key parameters in the electrolytic chlorine production process, such as salinity, current efficiency, actual sodium hypochlorite product concentration, and electrode state, cannot be directly measured by sensors, and these state variables directly affect the system's control performance and product quality, it is necessary to infer the system's internal state based on observable process parameters using a state estimation algorithm. This module organically integrates the mechanistic knowledge of the digital twin model in module 102 with the real-time observation data from module 101, achieving optimal estimation of the system state through an extended Kalman filter algorithm.

[0092] The core of the state estimation algorithm lies in establishing a state-space model of the system, which is directly discretized based on the digital twin model constructed in module 102. The state vector includes unmeasurable state variables, such as seawater salinity, current efficiency, electrode state, and the actual product concentration of sodium hypochlorite. Seawater salinity, as a fundamental parameter affecting the efficiency of the electrolysis reaction, directly influences the calculated current efficiency. Current efficiency reflects the effective conversion of electrical energy into chemical energy during electrolysis and is a crucial link between the electrochemical model and product formation. Electrode state describes the degree of scaling and aging on the electrode surface, directly affecting the efficiency of the electrolysis reaction and the long-term stability of the system. The actual product concentration of sodium hypochlorite is the final output of the system and is a core target variable that the control system needs to track.

[0093] The observation vector includes observed variables, namely, conductivity, cell voltage, and observed cell outlet temperature. Conductivity is strongly correlated with seawater salinity, providing key observational information for salinity estimation. Cell voltage reflects the electrical characteristics of the electrolysis process and is closely related to electrode state and electrolysis efficiency. The observed cell outlet temperature reflects the thermal effects of the electrolysis process and is directly related to electrolysis power and reaction intensity. These observed variables are acquired in real-time through the data processing system in module 101, providing reliable observational data input for the state estimation algorithm.

[0094] The state transition equations are established based on the dynamic characteristics of the digital twin model in Module 102, describing the evolution of the system state over time. The state transition for seawater salinity considers the slow changes in seawater supply, employing a first-order inertial element to describe its dynamic process, with the time constant reflecting the slow changes in seawater composition. The state transition for current efficiency is based on the deep learning efficiency calculation model in Module 102, considering the combined effects of temperature, salinity, and current density on efficiency. The state transition for electrode state describes the progressive aging process of the electrode, considering the cumulative effects of current density, operating time, and cleaning frequency on electrode performance. The state transition for the actual product concentration of sodium hypochlorite is directly based on the product formation equation of the electrochemical model in Module 102, reflecting the instantaneous response characteristics of the electrolysis reaction.

[0095] The observation equations establish the functional relationship between observed variables and state variables, directly adopting the corresponding sub-models in the digital twin model of Module 102. The conductivity observation equation, based on the salinity estimation model of Module 102, establishes a nonlinear relationship between conductivity and salinity and temperature. The cell voltage observation equation considers the electrochemical characteristics of the electrolysis reaction, linking cell voltage with state variables such as current efficiency, electrode state, and electrolysis current. The electrolytic cell outlet temperature observation equation directly adopts the temperature prediction results from the thermodynamic model of Module 102, establishing an energy balance relationship between temperature and electrolysis power and heat dissipation loss.

[0096] The implementation of the Extended Kalman Filter (EKF) algorithm includes iterative processes of prediction and update steps. The prediction step, based on the system's state transition equations and the previous time-ahead posterior state estimate, calculates the current time-ahead state prior estimate and covariance prior estimate. The state prior estimate is obtained by applying the state transition function to the previous time-ahead state posterior estimate and the current control input vector, reflecting the natural evolution of the system state. The covariance prior estimate propagates the previous time-ahead covariance posterior estimate through the state transition matrix and incorporates the effects of process noise, reflecting the uncertainty of state prediction.

[0097] The update step uses the current observation data to correct the prior estimate of the state, obtaining a more accurate posterior estimate. The calculation of the Kalman gain matrix comprehensively considers the uncertainty of state prediction and the reliability of observation data, achieving an optimal trade-off between the two. When the reliability of the observation data is high, the Kalman gain increases, relying more on observation information for state correction. When the accuracy of the state prediction is high, the Kalman gain decreases, preserving the prediction result more. The posterior state estimate incorporates the weighted observation residuals into the prior state estimate through the Kalman gain matrix, achieving an optimal fusion of observation information and model prediction.

[0098] Key technical aspects of the algorithm implementation include the calculation of the Jacobian matrix, the setting of the noise covariance matrix, and the guarantee of numerical stability. The Jacobian matrix is ​​obtained by numerically differentiating the state transition function and the observation function, reflecting the system's local linearization characteristics. Setting the process noise covariance matrix requires comprehensive consideration of model uncertainties and the impact of external disturbances, determining appropriate values ​​through historical data statistical analysis and engineering experience. The observation noise covariance matrix is ​​set based on the sensor's measurement accuracy and the filtering effect of the data processing module, ensuring the algorithm's accurate assessment of the observation data quality.

[0099] To ensure the numerical stability of the algorithm, square root filtering is employed to avoid numerical degradation of the covariance matrix. Simultaneously, state constraint processing is implemented to ensure that the estimated state variables remain within a physically reasonable range. When an anomaly occurs in the estimation result, a fault detection and isolation mechanism is activated, automatically switching to a backup estimation strategy to guarantee the continuous and stable operation of the system.

[0100] The state variables estimated in this module include seawater salinity, current efficiency, electrode state, and the actual product concentration of sodium hypochlorite. This state information is transmitted in real time to the multivariate predictive control algorithm in module 104 as the basic state input for control decisions. Simultaneously, the state estimation results provide crucial internal state monitoring information to the safety monitoring module in module 105, supporting the implementation of predictive safety control strategies. Furthermore, the accuracy evaluation data of the state estimation is fed back to the online learning module in module 106 for continuous optimization of the digital twin model parameters and performance improvement of the state estimation algorithm.

[0101] Predictive control module 104, based on the digital twin model and state vector, constructs an objective function, solves the objective function to obtain control commands, specifically as follows: Figure 4 As shown.

[0102] Based on the digital twin model and state vector, this module adopts a multivariate predictive control algorithm to optimize the control of the electrolytic chlorine production process. Through a rolling optimization strategy, under the premise of meeting safety and process constraints, it achieves accurate tracking of sodium hypochlorite product concentration, minimization of system energy consumption, protection of electrode state, and stability of control variable adjustment. The optimal control command is generated and transmitted to module 105 for safety verification and fault handling.

[0103] The predictive control module plays a core decision-making role in the entire intelligent control system. It organically combines the predictive capabilities of the digital twin model in module 102 with the state perception of the state estimation in module 103, and achieves the global optimal operation of the system through advanced optimization control algorithms. This module receives key state variables provided by module 103, such as seawater salinity, current efficiency, electrode state, and the actual product concentration of sodium hypochlorite. Combining these with the predictions of the system's future behavior from the digital twin model in module 102, it solves a multi-objective optimization problem in the prediction time domain, generating an optimal control strategy that balances economy, safety, and stability.

[0104] The core of the multivariate predictive control algorithm lies in establishing a comprehensive performance objective function that holistically considers multiple control objectives in the electrolytic chlorination process. Sodium hypochlorite product concentration tracking performance is the primary control objective, ensuring product quality meets downstream user requirements. Precise tracking control is achieved by minimizing the sum of squares of the deviations between the actual product concentration and the preset target concentration. System energy consumption optimization, as an economic objective, reduces operating costs by minimizing total power consumption in the prediction time domain. This power consumption mainly includes the energy consumption of equipment such as electrolysis power, seawater booster pump power, and blower power. Electrode condition protection, as an equipment maintenance objective, avoids excessive electrode aging by optimizing current density distribution and operating conditions, extending equipment lifespan and reducing maintenance costs. Control variable adjustment stability, as an operational stability objective, avoids frequent adjustments that could impact equipment and processes by limiting the range of changes in control variables, ensuring stable system operation.

[0105] The objective function is to minimize the sum of performance indicators at each time point in the prediction time domain. The performance indicators include the product of the preset concentration tracking weight coefficient and the square of the difference, the product of the preset energy consumption optimization weight coefficient and the predicted power, the product of the preset electrode protection weight coefficient and the electrode state, and the product of the preset control stability weight coefficient and the square of the L2 norm of the control variable increment vector. Among them, the squared difference is equal to the square of the difference between the actual product concentration and the preset target concentration; the prediction time domain length is the time range of the multivariate predictive control algorithm; the preset target concentration is the desired product concentration; the predicted power is equal to the product of the cell voltage and the electrolysis current setpoint; the control variable increment vector is the change of the control variable between adjacent time points; the preset concentration tracking weight coefficient is the importance of the product concentration tracking performance, with a default value of 100, which is determined through control performance tuning; the preset energy consumption optimization weight coefficient is the importance of the energy consumption optimization performance, with a default value of 1, which is determined through economic benefit analysis; the preset electrode protection weight coefficient is the importance of the electrode protection performance, with a default value of 10, which is determined through equipment maintenance cost analysis; and the preset control stability weight coefficient is the importance of the control stability performance, with a default value of 0.1, which is determined through control stability analysis.

[0106] The objective function is mathematically expressed using a weighted multi-objective optimization approach, with each performance indicator reflecting its importance in the overall control objective through corresponding weight coefficients. The concentration tracking weight coefficient is set to a relatively large value to reflect the primary role of product quality control and ensure stable product concentration output under any operating conditions. The energy consumption optimization weight coefficient is determined based on electricity cost and economic benefit analysis, aiming to minimize energy consumption while maintaining product quality. The electrode protection weight coefficient is set based on equipment maintenance cost and replacement cycle analysis, balancing current production efficiency with long-term equipment health. The control variable adjustment stability weight coefficient is set to a relatively small value to ensure operational stability without affecting the main control objective.

[0107] The selection of control variables directly affects the system's control performance and operational flexibility. The electrolysis current setpoint, as a core control variable, directly determines the intensity of the electrolysis reaction and the product formation rate. Adjusting the electrolysis current allows for rapid response to product concentration tracking requirements. The seawater flow rate setpoint, achieved by adjusting the frequency converter of the seawater booster pump, affects the residence time and product concentration in the electrolyzer and plays a crucial role in system temperature control. The seawater booster pump frequency setpoint controls the pump's operating frequency, coordinating with the seawater flow rate setpoint to achieve precise flow control. The acid mist absorption fan speed setpoint controls the ventilation intensity of the acid mist absorption system, primarily used for hydrogen concentration control and system safety protection.

[0108] The setting of constraints ensures that the control strategy is implemented within a safe and feasible range, including safety constraints, process constraints, and control constraints. Safety constraints are set based on the system's safe operation requirements and are hard conditions that the control strategy must strictly meet. Safety constraints include electrolyzer outlet temperature constraints, hydrogen concentration constraints, flow rate lower limit constraints, and voltage upper limit constraints. Electrolyzer outlet temperature constraints prevent system overheating that could lead to equipment damage or safety accidents. This constraint includes two levels: soft constraints and hard constraints. Soft constraints serve as penalty terms in the optimization objective, while hard constraints serve as absolute boundary conditions that cannot be violated. The preset upper limit for the soft constraint on the electrolyzer outlet temperature is 38 degrees Celsius, with a default value of 38 degrees Celsius. This value is determined based on the equipment's normal operating temperature range and thermodynamic safety margin analysis, ensuring that control measures are taken in advance when the system temperature approaches the danger zone. The preset upper limit for the hard constraint on the electrolyzer outlet temperature is 42 degrees Celsius, with a default value of 42 degrees Celsius. This value is determined based on the temperature resistance limit of the equipment materials and safety specifications, serving as the absolute temperature boundary for emergency system shutdown. Hydrogen concentration constraints prevent hydrogen accumulation to the explosive limit. Active safety control is achieved by comparing the predicted hydrogen concentration with a preset safety threshold. The preset upper limit for hydrogen concentration constraints is 25% of the lower explosive limit (LEL), with a default value of 1% by volume. This is determined based on the product of the LLE by volume (4% of the LLE) and a safety factor of 4, ensuring that the hydrogen concentration remains within a safe range. Seawater flow rate constraints ensure sufficient cooling and reaction medium in the electrolyzer, preventing equipment damage due to excessively low flow rates. The preset lower limit for seawater flow rate constraints is 30% of the rated flow rate, with a default value of 0.3 cubic meters per hour. This is determined based on the minimum cooling requirements and reaction kinetics of the electrolyzer, guaranteeing the basic reaction conditions for the electrolysis process. Cell voltage upper limit constraints prevent excessive voltage from damaging electrodes or causing excessive energy consumption. The preset upper limit for cell voltage constraints is 110% of the design voltage, with a default value of 5.5 volts. This is determined based on the electrode material's withstand capability and energy consumption economics analysis, avoiding irreversible damage to the electrodes caused by excessive voltage.

[0109] Process constraints are set based on the normal operation requirements of the electrolytic chlorine production process to ensure the stability of the production process and product quality. These constraints include sodium hypochlorite storage tank level constraints, electrolysis current range constraints, and seawater booster pump frequency range constraints. The sodium hypochlorite storage tank level constraints ensure continuous product storage, preventing overflow due to excessively high levels or disruption to product supply due to excessively low levels. The preset upper limit for the sodium hypochlorite storage tank level constraints is 90% of the tank volume, with a default value of 90 cubic meters. This is determined based on the storage tank's safety volume and overflow protection requirements, ensuring storage safety and providing an appropriate safety margin. The preset lower limit for the sodium hypochlorite storage tank level constraints is 10% of the tank volume, with a default value of 10 cubic meters. This is determined based on the continuous product supply requirements and pump suction conditions, guaranteeing continuous product supply to downstream users. The electrolysis current range constraints are determined based on the design parameters of the electrolyzer and rectifier power supply, ensuring that the equipment operates within its rated operating range. The preset upper limit of the electrolysis current range constraint is 95% of the rated current of the rectifier power supply, with a default value of 950 amps. This is determined based on the overload protection of the rectifier power supply and the requirements for safe operation of the equipment, to avoid equipment damage caused by power supply overload. The preset lower limit of the electrolysis current range constraint is 10% of the rated current of the rectifier power supply, with a default value of 100 amps. This is determined based on the minimum starting current of the electrolysis reaction and the product quality requirements, to ensure that the electrolysis reaction can proceed normally. The frequency range constraint of the seawater booster pump is determined based on the pump's performance curve and the adjustment range of the frequency converter, to ensure the pump's efficient and stable operation. The preset upper limit of the frequency range constraint of the seawater booster pump is 95% of the rated frequency of the frequency converter, with a default value of 47.5 Hz. This is determined based on the frequency converter's overload protection and the pump's mechanical strength requirements, to avoid equipment damage caused by high-frequency operation. The preset lower limit of the frequency range constraint of the seawater booster pump is 20% of the rated frequency of the frequency converter, with a default value of 10 Hz. This is determined based on the pump's minimum stable operating frequency and the requirements for flow control accuracy, to ensure that the pump can operate stably and provide accurate flow control.

[0110] Control constraints are set based on the execution capability and operational safety requirements of the control system to ensure the executability of control commands and the stability of the system. These constraints include increment limits and range constraints for control variables. Increment limits prevent drastic changes in control variables from impacting the system, achieving smooth control by limiting the maximum variation of control variables between adjacent time points. The preset upper limit for the increment limit of the electrolysis current setpoint is 5% of the rated current, with a default value of 50 amperes per minute. This is determined based on the dynamic response characteristics of the electrolysis reaction and the thermal inertia requirements of the equipment, avoiding temperature shocks and electrode damage caused by rapid current changes. The preset upper limit for the increment limit of the seawater flow rate setpoint is 10% of the rated flow rate, with a default value of 0.1 cubic meters per hour per minute. This is determined based on the response characteristics of the flow control system and process stability requirements, ensuring smooth flow regulation and system stability. The preset upper limit for the increment limit of the seawater booster pump frequency setpoint is 5% of the rated frequency, with a default value of 2.5 Hz per minute. This is determined based on the inverter's adjustment characteristics and the pump's mechanical response requirements, avoiding mechanical shocks to the pump caused by rapid frequency changes. The preset upper limit for the incremental limit of the acid mist absorption fan speed setpoint is 10% of the rated speed, with a default value of 100 rpm. This limit is determined based on the fan's adjustment characteristics and the stability requirements of the ventilation system, ensuring the smoothness of ventilation control. Control variable range constraints ensure that all control variables are within the adjustable range of the actuator, avoiding performance degradation caused by control saturation. The preset range for the electrolysis current setpoint is the adjustable range of the rectifier power supply, with a default value of 100 amps to 950 amps, determined based on the rectifier power supply specifications and electrolysis process requirements. The preset range for the seawater flow rate setpoint is the adjustable range of the seawater booster pump, with a default value of 0.3 cubic meters per hour to 3 cubic meters per hour, determined based on the pump performance curve and process flow requirements. The preset range for the seawater booster pump frequency setpoint is the adjustable range of the frequency converter, with a default value of 10 Hz to 47.5 Hz, determined based on the frequency converter specifications and pump operating characteristics. The preset range of the acid mist absorption fan speed setting is the adjustable range of the fan, with a default value of 300 rpm to 1500 rpm, determined based on the fan performance curve and ventilation requirements.

[0111] The optimization solution process transforms the multivariate predictive control problem into a standard quadratic programming problem. The objective function is transformed into a quadratic form through Taylor expansion and linearization, and the constraints are transformed into linear inequality constraints through linearization. The Hessian matrix represents the coefficients of the quadratic terms of the objective function, reflecting the coupling relationship between the control variables and the convexity of the optimization problem. The gradient vector represents the coefficients of the linear terms of the objective function, reflecting the degree of influence of each control variable on the objective function under the current operating conditions. The constraint matrix and constraint vectors uniformly express all constraints in matrix inequality form, facilitating the implementation of the numerical solution algorithm.

[0112] The numerical solution algorithm employs mature quadratic programming methods such as the interior-point method or the effective set method, which are characterized by fast convergence speed and good numerical stability. Real-time requirements are considered during the solution process, and a warm-start technique is used to utilize the optimization results from the previous time step as the initial value for the current optimization, thereby improving solution efficiency. Simultaneously, feasibility checks are implemented; when constraints become infeasible, constraint parameters are automatically adjusted or a backup control strategy is switched to ensure the continuous and stable operation of the system.

[0113] The rolling optimization strategy embodies the core idea of ​​predictive control. It solves a finite-time optimization problem within each control cycle, implementing only the first setpoint for the control variable, and then resolving the optimization problem based on the new state information. This strategy effectively handles the effects of model uncertainty and external disturbances, ensuring the robustness of control performance through a feedback correction mechanism. The choice of prediction time domain length requires a balance between control performance and computational complexity. Too short a prediction time domain may lead to poor control performance, while too long a prediction time domain increases the computational burden and may reduce the accuracy of model predictions.

[0114] The optimal control commands generated by this module include multiple control variables, such as the electrolysis current setpoint, seawater flow rate setpoint, seawater booster pump frequency setpoint, and acid mist absorption fan speed setpoint. These control commands are transmitted to module 105 in real time for safety verification and fault handling, ensuring that all control variable adjustments meet safety constraints before being issued to the field for execution. Simultaneously, control performance evaluation data, including objective function values, constraint violations, and solution convergence information, are fed back to the online learning module of module 106 for adaptive adjustment of control algorithm parameters and continuous optimization of the control strategy. Furthermore, the intermediate calculation results of predictive control provide module 105 with predictive information about the future state of the system for safety monitoring, supporting the implementation of predictive safety control strategies.

[0115] The safety monitoring module 105 performs safety verification on the control commands and executes safety protection measures based on the predicted value of the electrolytic cell outlet temperature and the state vector.

[0116] Based on optimal control commands and state vectors, this module constructs a multi-layered safety protection system. Through the synergistic effect of predictive safety control and hardware safety interlocks, it achieves comprehensive safety assurance for the electrolytic chlorine production process, ensuring that the system remains in a safe and controllable state under complex and ever-changing operating conditions. Safety protection measures include hydrogen concentration safety control strategies and temperature safety control strategies.

[0117] The safety monitoring module employs a predictive safety control strategy based on a digital twin model. It fully utilizes the predictive capabilities and state vectors of the digital twin model provided by module 102 to achieve early identification and proactive prevention of safety risks. Predictive safety control continuously monitors the changing trends of key system safety parameters and, combined with the digital twin model's predictions of future states, takes preventative control measures before safety risks evolve into actual hazards.

[0118] The system establishes a safety status assessment mechanism, comprehensively considering multiple safety indicators such as hydrogen concentration, electrolyzer temperature, electrode condition, and electrical parameters, and constructs a safety status vector for real-time assessment. The safety status assessment results are divided into three levels: safe, warning, and hazardous, each corresponding to different control strategies and response measures. Under the safe level, the system operates normally according to the optimized control instructions of module 104; under the warning level, preventative control measures are initiated and control constraints are adjusted; under the hazardous level, emergency control procedures are executed and preparations for safe shutdown are made.

[0119] Predictive safety control is deeply integrated with the multivariate predictive control algorithm of module 104. By dynamically adjusting control constraints and objective function weights, safety control requirements are embedded into the optimization control process. When predictive safety control detects a potential safety risk, it automatically tightens the constraints of relevant control variables and increases the weight of the safety objective in the optimization objective function, ensuring that control decisions prioritize meeting safety requirements.

[0120] Hydrogen, as a byproduct of the chlorine electrolysis process, has a concentration whose control directly affects the inherent safety of the system. The hydrogen concentration safety control strategy establishes a tiered safety control system based on dynamic monitoring and prediction of hydrogen concentration. This system utilizes various methods, including fan speed regulation, electrolysis current control, and ventilation system management, to ensure that the hydrogen concentration remains within a safe range.

[0121] Under normal operating conditions, the hydrogen concentration control target is set below a preset normal hydrogen threshold, providing sufficient safety margin for the system. When the detected hydrogen concentration exceeds the preset allowable hydrogen threshold, the system enters an early warning control state. At this time, the predictive safety control algorithm initiates hydrogen concentration trend analysis, calculating the future development trend of hydrogen concentration based on the current concentration change rate and digital twin model prediction. The default value of the preset normal hydrogen threshold is 60% of the preset upper limit of the hydrogen concentration constraint, and the default value of the preset allowable hydrogen threshold is 70% of the preset upper limit of the hydrogen concentration constraint.

[0122] Under early warning control conditions, the system employs a coordinated control strategy to simultaneously adjust multiple control variables. The speed of the acid mist absorption fan increases according to a proportional control law, and the fan speed adjustment is determined based on the hydrogen concentration deviation and concentration change rate to ensure that the ventilation capacity matches the hydrogen generation rate. The electrolysis current adopts a gradual load reduction strategy to avoid sudden current changes impacting the electrolysis process. The current adjustment rate is controlled to not exceed 5% of the rated current per minute, ensuring both safety and maintaining production stability.

[0123] When the hydrogen concentration continues to rise to 85% of the maximum permissible value, the system enters a hazardous control state and initiates more stringent control measures. At this time, the system forcibly reduces the electrolysis current to the safe operating lower limit, while simultaneously increasing the speed of the acid mist absorption fan to the maximum permissible value and activating backup ventilation equipment to enhance ventilation. Under hazardous control conditions, the system suspends receiving optimized control commands from module 104 and operates entirely according to the safety control logic until the hydrogen concentration drops to a safe range.

[0124] As a last line of defense, the hardware safety interlock triggers an emergency shutdown procedure when the hydrogen concentration reaches the maximum permissible level. This procedure cuts off the electrolysis power supply, activates the emergency ventilation system, and simultaneously triggers an audible and visual alarm to alert operators to take emergency measures. After the hardware interlock activates, the system enters a safe shutdown state and can only be restarted after safety confirmation and system checks.

[0125] Electrolytic cell temperature control is a crucial aspect of ensuring the safe and stable operation of the electrolysis process. Excessively high temperatures not only affect electrolysis efficiency but can also lead to equipment damage and safety accidents. The temperature safety control strategy establishes a temperature safety control system based on temperature prediction and multivariate coordination. This system utilizes methods such as seawater flow regulation, electrolysis current control, and heat dissipation system management to achieve precise temperature control of the electrolytic cell.

[0126] Temperature safety control fully utilizes the predictive capabilities of the thermodynamic model in the digital twin model of module 102 to predict future trends in the electrolyzer outlet temperature based on current operating parameters. Predictive temperature control not only considers current temperature measurements but, more importantly, the temperature development trends predicted by the thermodynamic model, achieving proactive and forward-looking temperature control.

[0127] Under normal operating conditions, the target temperature control for the electrolyzer outlet is set below a preset safety threshold, providing ample adjustment space for temperature control. When the predicted outlet temperature exceeds the preset warning threshold, the system activates a warning control mode, achieving temperature control by coordinating the adjustment of seawater flow rate and electrolysis current. Seawater flow rate adjustment employs a feedforward and feedback control strategy. Feedforward control adjusts the flow rate in advance based on temperature change trends predicted by a thermodynamic model, while feedback control performs fine adjustment based on actual temperature measurements. The default preset safety threshold is 35 degrees Celsius, and the default preset warning threshold is 37 degrees Celsius.

[0128] When the predicted outlet temperature of the electrolyzer exceeds the preset emergency threshold, the system enters emergency control mode and adopts more aggressive control measures. At this time, the system prioritizes temperature safety, temporarily sacrificing some economic indicators by increasing the seawater flow rate to the maximum allowable value, while appropriately reducing the electrolysis current to decrease heat generation. In emergency control mode, the system also activates auxiliary cooling equipment, such as the cooling water circulation system and cooling fans, to enhance the system's heat dissipation capacity. The default preset emergency threshold is 39 degrees Celsius.

[0129] The hardware temperature protection interlock is set at 42 degrees Celsius. When the observed temperature at the electrolytic cell outlet reaches this threshold, the PLC system immediately executes a protective shutdown, cutting off the electrolysis power supply and maximizing seawater flow for cooling to prevent overheating and damage to the equipment. After the temperature interlock is activated, the system must wait for the temperature to drop to a safe range and for the equipment to be inspected and confirmed to be free of abnormalities before it can be restarted.

[0130] This module interacts with state vectors and control commands in real time to form a closed-loop feedback mechanism for safety control. Safety control decision results are fed back to module 104 in real time for dynamic adjustment of control constraints, ensuring that optimized control always operates within safety constraints. Simultaneously, safety status assessment information provides safety-related performance indicators for the online learning of module 106, supporting continuous optimization of safety control strategies.

[0131] The online learning module 106 receives accuracy evaluation data of the state vector, real-time evaluation indicators of the control performance of module 104, and execution effect information of safety control of module 105. This module constructs an adaptive learning mechanism to ensure that the entire intelligent control system always maintains the optimal operating state and control performance through online updates of model parameters, dynamic optimization of control strategies, and continuous improvement of system performance.

[0132] This module receives accuracy assessment data of the state vector, real-time evaluation indicators of the control performance of module 104, and execution effect information of the safety control of module 105. It constructs an adaptive learning mechanism to ensure that the entire intelligent control system always maintains optimal operating state and control performance through online updating of model parameters, dynamic optimization of control strategies, and continuous improvement of system performance.

[0133] The core function of the online learning module is to achieve adaptive updates of digital twin model parameters to address the impact of factors such as changes in operating conditions, equipment aging, and environmental disturbances on model accuracy during actual production. The adaptive learning mechanism, based on state vector error analysis and the control performance evaluation of module 104, identifies the root causes of model parameter deviations and employs targeted parameter update strategies.

[0134] The salinity estimation model employs a recursive least squares algorithm for parameter learning, combined with a forgetting factor mechanism to handle time-varying characteristics. When the salinity estimation error exceeds a preset threshold, the online learning module initiates a parameter update procedure for the salinity estimation model. This preset threshold, with a default value of 3%, is determined based on a balance between salinity estimation accuracy requirements and system stability. The parameter update process first analyzes the statistical characteristics of the estimation error to determine whether the error is caused by random noise or by model parameter shifts. If the error exhibits systematic bias characteristics, the parameter update algorithm is activated to correct the model parameters using the latest measurement data.

[0135] Online learning of the efficiency calculation model is more complex, requiring the design of an incremental learning strategy that incorporates the characteristics of deep learning models. When new product concentration test data is obtained, the online learning module first calculates the deviation between the actual current efficiency and the model's predicted value, analyzing the distribution characteristics and trends of the deviation. If the deviation exceeds the preset normal fluctuation range, the system initiates incremental updates to the model parameters, using a mini-batch gradient descent algorithm to fine-tune the neural network weights, thus maintaining the model's memory of historical knowledge while adapting to new operating conditions. The default value for this preset normal fluctuation range is 2%, determined based on the accuracy of current efficiency measurement and the model's predictive capability.

[0136] The learning of thermodynamic model parameters mainly focuses on the time-varying characteristics of physical parameters such as heat transfer coefficient and heat dissipation properties. By comparing the electrolyzer outlet temperature predicted by the thermodynamic model with the actual measured value, deviations in heat transfer parameters are identified. Parameter updates employ a physics-constrained optimization algorithm to ensure that the updated parameters still satisfy the fundamental laws of thermodynamics and the constraints of the equipment's physical properties.

[0137] The online learning module not only focuses on updating model parameters, but more importantly, optimizes control strategies and parameters based on feedback information from control performance. By analyzing the execution effect of the multivariate predictive control in module 104, including the tracking accuracy of the control target, the stability of the control variables, and the degree to which constraints are satisfied, areas for improvement in control performance are identified.

[0138] The prediction and control time domain parameters of predictive control are also adaptively adjusted according to changes in the system's dynamic characteristics. By analyzing the system's dynamic response characteristics and disturbance characteristics, the optimal prediction time domain length is dynamically determined, ensuring prediction accuracy while avoiding excessive computational burden. The selection of the control time domain comprehensively considers control performance and real-time computation requirements, achieving the best balance between performance and efficiency.

[0139] The online learning module establishes a comprehensive performance evaluation system to regularly assess the operational effectiveness of the entire intelligent control system, including multiple dimensions such as product quality stability, energy consumption efficiency, equipment operating status, and safety control effectiveness. The performance evaluation results are not only used for model and control parameter optimization but also provide decision support for adjusting system operation strategies.

[0140] System performance evaluation employs a multi-level indicator system, including immediate performance indicators, short-term performance indicators, and long-term performance indicators. Immediate performance indicators reflect the system's current operating status, such as control error, energy consumption level, and safety margin. Short-term performance indicators evaluate the system's overall performance within a certain time window, such as product quality pass rate, equipment utilization rate, and failure frequency. Long-term performance indicators focus on the system's sustainable operating capability, such as equipment lifespan, maintenance costs, and overall economic benefits.

[0141] Based on the performance evaluation results, the online learning module generates system optimization suggestions, including suggestions for adjusting operating parameters, optimizing maintenance plans, and improving control strategies. These optimization suggestions not only provide feedback to relevant control modules for real-time adjustments but also provide a scientific basis for production management decisions.

[0142] The learning outcomes and optimization suggestions from this module are passed to other modules through standardized interfaces, forming a system-level collaborative optimization mechanism. Updated model parameters are passed to module 102 in real time to improve the accuracy of the digital twin model, optimized control parameters are fed back to module 104 for continuous improvement of control performance, and optimization suggestions for safety control strategies are provided to module 105 to enhance safety protection capabilities.

[0143] The optimization scheduling module 107 comprehensively utilizes the electrode status provided by module 103, the safety status assessment results of module 105, and the performance optimization suggestions of module 106. This module constructs an intelligent pickling scheduling system. Through precise assessment of electrode status, optimized selection of pickling timing, safety control of the pickling process, and quantitative evaluation of pickling effect, it achieves optimal coordination between electrode maintenance and production operation, ensuring that the system maintains the best working condition of the electrodes while maintaining efficient production.

[0144] Electrode pickling is a crucial step in maintaining the efficient operation of the electrolysis system, and its timing directly impacts production continuity and economic benefits. This module, based on the electrode status provided by module 103, combined with the safety status assessment from module 105 and performance optimization suggestions from module 106, establishes a multi-dimensional pickling triggering mechanism and a comprehensive optimization decision-making system.

[0145] The pickling requirement assessment comprehensively considers three dimensions: electrode contamination level, electrolytic performance degradation, and system operational safety. Electrode contamination level is monitored in real-time by the electrode state estimator in module 103. When the electrode surface contamination index exceeds a preset threshold, it indicates that accumulated contaminants on the electrode surface have begun to affect electrolytic efficiency. This preset threshold has a default value of 0.8 and is determined based on experimental data on the impact of electrode contamination on electrolytic efficiency. Electrolytic performance degradation is assessed by the magnitude of current efficiency decrease and the degree of abnormal increase in cell voltage. When the current efficiency decreases by more than 5% compared to the baseline value or the cell voltage increases by more than 0.5 volts compared to the normal value, it indicates that electrode performance has significantly degraded. This preset 5% default value is determined based on an economic analysis of electrolytic efficiency, and the preset 0.5 volts default value is determined based on the abnormal cell voltage judgment criteria.

[0146] The decision-making process for optimizing pickling timing employs a multi-objective optimization method, comprehensively considering production benefit losses, pickling operation costs, and downtime impact costs. Production benefit losses primarily consist of reduced output and increased energy consumption due to decreased current efficiency, quantified by calculating the difference between current and standard efficiency and the corresponding economic losses. Pickling operation costs include direct costs such as acid consumption, labor input, and equipment wear and tear. Downtime impact costs consider the amortization of fixed costs during downtime and potential losses from market supply disruptions.

[0147] Pickling scheduling decisions also need to consider production planning constraints and system operation constraints. Production planning constraints require pickling operations to be scheduled during periods of lower production load to avoid affecting the delivery of important orders. System operation constraints include ensuring that the storage liquid level meets the supply demand during pickling, that fluctuations in downstream user demand do not exceed the storage buffer capacity, and that the pickling frequency does not exceed the equipment's capacity.

[0148] The safe and efficient execution of the pickling process requires close coordination with other modules. Before pickling starts, the safety monitoring system in module 105 conducts a comprehensive safety check to confirm that the hydrogen concentration is within a safe range, the electrolyzer temperature has dropped to an appropriate level, and all safety interlock systems are in normal working order. The predictive control system in module 104 gradually reduces the electrolysis current according to a predetermined program to avoid sudden current surges that could impact the equipment, while simultaneously adjusting seawater flow and other auxiliary system parameters to create optimal conditions for pickling.

[0149] The pickling process is divided into three stages: system preparation, pickling execution, and system recovery. Each stage has clear control objectives and safety requirements. The system preparation stage mainly involves the safe shutdown of the production system and the startup preparation of the pickling system, including the safe disconnection of the electrolysis power supply, confirmation of the connectivity of the pickling pipeline, preparation of the pickling solution, and startup testing of the circulation system. This stage requires ensuring that the storage system has sufficient liquid levels to meet the external supply needs during pickling, while simultaneously activating the backup supply plan to ensure that downstream user demand is not affected.

[0150] During the pickling process, the flow rate, temperature, and concentration of the pickling solution are strictly controlled according to process requirements. The pH and conductivity of the pickling solution are monitored online to assess the pickling effect and progress in real time. Throughout the pickling process, changes in the electrode surface condition are continuously monitored, and the contaminant removal effect is evaluated using methods such as electrochemical impedance spectroscopy to ensure that the pickling achieves the expected goals. Simultaneously, the pickling time is strictly controlled to avoid over-pickling and damage to the electrodes, generally kept within a preset range of two to four hours. This preset range of two to four hours is determined based on the electrode material characteristics and pickling process optimization experiments.

[0151] The system recovery phase includes thorough removal of the pickling solution, thorough rinsing of electrodes and pipelines, restarting of the electrolysis system, and online calibration of operating parameters. The cleaning process must ensure complete removal of the pickling solution to avoid residual acid adversely affecting subsequent electrolysis processes. The system restart is performed according to standard startup procedures, gradually increasing the electrolysis current to normal operating levels while closely monitoring all operating parameters to ensure the system quickly returns to its optimal operating condition.

[0152] After pickling is completed, this module establishes a comprehensive effect evaluation mechanism, evaluating the pickling effect through multi-dimensional indicators to provide data support for subsequent pickling strategy optimization. The effect evaluation is mainly conducted from three aspects: the degree of electrode performance recovery, system operational stability, and improvement in economic benefits.

[0153] The degree of electrode performance recovery is assessed by comparing key parameters such as current efficiency, tank voltage, and electrode impedance before and after pickling. Under normal circumstances, effective pickling should restore current efficiency to over 95% of the preset standard level, reduce tank voltage to the preset normal range, and effectively reduce electrode impedance. This preset standard level is determined based on the electrolysis system's design performance indicators, and this preset normal range is determined based on statistical analysis of historical equipment operating data. If the pickling effect is unsatisfactory, the causes need to be analyzed and pickling process parameters, such as acid concentration, pickling time, and circulation flow rate, adjusted.

[0154] System stability assessment focuses on the dynamic response characteristics and control performance of the system after pickling. By analyzing the smoothness of the system startup process, the regulation characteristics of control variables, and the disturbance response capability, the impact of pickling on the overall system performance is evaluated. High-quality pickling should not only restore electrode performance but also improve the system's dynamic characteristics, enhancing control accuracy and response speed.

[0155] The economic benefit assessment comprehensively considers the cost input and efficiency improvement of pickling, calculating the return on investment and economic benefits of pickling. By comparing energy consumption levels, product quality, and production efficiency before and after pickling, the economic value brought by pickling is quantified. Simultaneously, the impact of pickling on equipment lifespan and maintenance costs is evaluated, providing a basis for long-term optimization of the pickling strategy.

[0156] This module achieves optimal coordination between pickling operations and overall production through deep collaboration with other modules. Pickling decision information is transmitted in real time to module 104 for production plan adjustments and control strategy optimization; pickling status information is fed back to module 105 for key adjustments in safety monitoring; and pickling effect data is provided to module 106 for system performance evaluation and model parameter updates. This collaborative mechanism ensures that pickling operations not only effectively maintain electrode performance but also form an organic whole with the entire intelligent control system, achieving continuous optimization of system performance.

[0157] Through the collaborative operation of modules 101-107, this system achieves digital and intelligent control of the entire electrolytic chlorine production process. Module 101 provides a reliable data foundation, module 102 constructs an accurate digital twin model, module 103 achieves accurate state estimation, module 104 executes optimized predictive control, module 105 ensures safe system operation, module 106 continuously optimizes system performance, and module 107 intelligently schedules acid washing operations, forming a complete closed-loop control system that improves the operating efficiency, product quality, and safety of the electrolytic chlorine production system.

[0158] Through the above technical solution, this embodiment successfully realizes intelligent control of the entire electrolytic chlorine production process, improving the system's control accuracy, energy efficiency, and economic benefits while ensuring safe production.

[0159] The embodiments of the present invention have been described above. However, the embodiments are not limited to the specific implementation methods described above. The specific implementation methods described above are merely illustrative and not restrictive. Those skilled in the art can make more equivalent embodiments under the guidance of the present embodiments, and all of them are within the protection scope of the present embodiments.

Claims

1. A fully intelligent control system for the electrolytic chlorine production process based on digital twins, characterized in that: include: The data processing module is used to collect and preprocess process parameters for the entire chemical production process in real time. The digital twin module constructs a digital twin model based on the process parameters. The digital twin model calculates the current efficiency, actual product concentration, predicted value of electrolytic cell outlet temperature, electrode state, and salinity. The state estimation module constructs and updates the state vector based on the process parameters, current efficiency, actual product concentration, electrode state, and salinity. The predictive control module constructs an objective function based on the digital twin model and the state vector, and solves the objective function to obtain control commands; The safety monitoring module performs safety verification on control commands and executes safety protection measures based on the predicted value of the electrolytic cell outlet temperature and the state vector.

2. The intelligent control system for the entire electrolytic chlorine production process based on digital twins according to claim 1, characterized in that, Digital twin models include electrochemical models, thermodynamic models, electrode state models, and salinity estimation models, among which: The actual product concentration is calculated by the electrochemical model. The actual product concentration is the product of the current efficiency and the electrolysis current, and then divided by the product of the preset electron transfer number, the preset Faraday constant and the seawater flow rate. The current efficiency is calculated by the efficiency calculation model. The thermodynamic model calculates the predicted value of the electrolyzer outlet temperature using the energy balance equation; The electrode state is calculated by the electrode state model. The electrode state at the next moment is the product of the preset state attenuation coefficient and the current electrode state, plus the product of the preset scale growth coefficient and the voltage term. The voltage term is the maximum value between 0 and the voltage difference. The voltage difference is the difference between the current tank voltage and the reference voltage. The salinity estimation model uses a multiple linear regression method to calculate the salinity of seawater. The salinity is calculated by adding a preset second constant term to the product of the conductivity coefficient and conductivity, the preset temperature coefficient and temperature, and the preset interaction term coefficient of conductivity and temperature to the product of conductivity and temperature.

3. The intelligent control system for the entire electrolytic chlorine production process based on digital twins according to claim 2, characterized in that, The reference voltage is the sum of the following: a preset first constant term, the product of the first-order coefficient of the preset salinity and salinity, the product of the first-order coefficient of the preset temperature and temperature, the product of the first-order coefficient of the preset current and current, the product of the quadratic coefficient of the preset salinity and the square of salinity, the product of the quadratic coefficient of the preset temperature and the square of temperature, the product of the quadratic coefficient of the preset current and the square of current, the product of the interaction coefficient of the preset salinity and temperature and the product of salinity and temperature, the product of the interaction coefficient of the preset salinity and current and the product of the interaction coefficient of the preset temperature and current and the product of temperature and current.

4. The intelligent control system for the entire electrolytic chlorine production process based on digital twins according to claim 2, characterized in that, The energy balance equation is: the product of seawater density, seawater specific heat capacity, and seawater flow rate, multiplied by the difference between the predicted outlet temperature and the inlet temperature of the electrolytic cell, equals the electrolysis power minus the heat dissipation loss power; the electrolysis power is the product of cell voltage and electrolysis current, and the heat dissipation loss power is the product of the preset total heat transfer coefficient and the preset heat exchange area, multiplied by the difference between the average temperature and the ambient temperature, where the average temperature is the arithmetic mean of the predicted inlet temperature and the outlet temperature of the electrolytic cell.

5. The intelligent control system for the entire electrolytic chlorine production process based on digital twins according to claim 1, characterized in that, The control commands received include: The objective function is to minimize the sum of performance indicators at each time point in the prediction time domain. The performance indicators include the product of the preset concentration tracking weight coefficient and the square of the difference, the product of the preset energy consumption optimization weight coefficient and the predicted power, the product of the preset electrode protection weight coefficient and the electrode state, and the product of the preset control stability weight coefficient and the square of the L2 norm of the control variable increment vector. Here, the square of the difference is equal to the square of the difference between the actual product concentration and the preset target concentration setting value, the predicted power is equal to the product of the cell voltage and the electrolysis current setting value, and the control variable increment vector is the change of the control variable between adjacent time points. A multivariable predictive control algorithm is used to solve the objective function under constraints to obtain control commands.

6. The intelligent control system for the entire electrolytic chlorine production process based on digital twins according to claim 5, characterized in that, The control command includes multiple control variables, including the electrolysis current setpoint, seawater flow rate setpoint, seawater booster pump frequency setpoint, and acid mist absorption fan speed setpoint.

7. The intelligent control system for the entire electrolytic chlorine production process based on digital twins according to claim 5, characterized in that, The constraints include safety constraints, process constraints, and control constraints. Among them, safety constraints include electrolyzer outlet temperature constraints, hydrogen concentration constraints, lower flow rate constraints, and upper voltage constraints; process constraints include sodium hypochlorite storage tank level constraints, electrolysis current range constraints, and seawater booster pump frequency range constraints; and control constraints include control variable increment limits and control variable range constraints.

8. The intelligent control system for the entire electrolytic chlorine production process based on digital twins according to claim 1, characterized in that, The safety protection measures include a hydrogen concentration safety control strategy and a temperature safety control strategy. The hydrogen concentration safety control strategy achieves graded control of hydrogen concentration through fan speed adjustment and electrolysis current control. Under normal operating conditions, the hydrogen concentration control target is set to be less than a preset normal hydrogen threshold. The temperature safety control strategy is based on a thermodynamic model. Under normal operating conditions, the electrolyzer outlet temperature control target is set to be less than a preset safety threshold.

9. The intelligent control system for the entire electrolytic chlorine production process based on digital twins according to claim 1, characterized in that, The state estimation module uses the extended Kalman filter algorithm to update the state vector. The state vector includes state variables, such as salinity, current efficiency, electrode state, and actual product concentration.

10. The intelligent control system for the entire electrolytic chlorine production process based on digital twins according to claim 1, characterized in that, The process parameters include flow rate parameters, electrical parameters, temperature parameters, pressure parameters, liquid level parameters, water quality parameters, and safety monitoring parameters. Among them, the flow rate parameters include seawater flow rate and sodium hypochlorite product delivery flow rate; the electrical parameters include electrolysis current and cell voltage; the temperature parameters include observed values ​​of the electrolytic cell inlet temperature and electrolytic cell outlet temperature; the pressure parameters include the electrolytic cell inlet pressure; the liquid level parameters include the liquid level in the sodium hypochlorite storage tank, the liquid level in the pickling tank, and the liquid level in the concentrated acid tank; the water quality parameters include conductivity and pH value; and the safety monitoring parameters include hydrogen concentration, equipment vibration signals, and electrical fault signals.

Images