A system for monitoring, evaluating and optimizing energy flow and carbon flow of an aluminum electrolytic production process

Abstract

The application discloses an energy flow and carbon flow monitoring, evaluation and optimization system of an aluminum electrolysis production process, and belongs to the technical field of metallurgical industry energy saving and emission reduction and intelligent control. System design steps include: designing the edge layer of the aluminum electrolysis production process energy flow and carbon flow control system, inputting the data collected by the edge layer into the cloud layer through the communication layer, and performing analysis operation on the energy flow-carbon flow coupling model; constructing an improved multi-objective genetic optimization algorithm in the cloud layer, and obtaining the optimal process parameter combination through the optimization objective function; dynamically adjusting, closed-loop feedback and abnormity processing operation through the execution mechanism of the edge layer; and comparing and analyzing the aluminum electrolysis production process energy flow and carbon flow control system, and performing economic analysis. The application realizes real-time monitoring of key parameters of the aluminum electrolysis process through multi-source data fusion technology, establishes an energy efficiency-carbon emission correlation evaluation system in combination with big data analysis, and generates an electrolytic cell process parameter dynamic adjustment strategy based on a self-adaptive optimization algorithm.

Description

Technical Field

[0001] This invention belongs to the field of energy conservation, emission reduction and intelligent control technology in the metallurgical industry, and specifically relates to a system for monitoring, evaluating and optimizing energy flow and carbon flow in aluminum electrolysis production process. Background Technology

[0002] Aluminum electrolysis, as a typical process industry with high energy consumption and high carbon emissions, accounts for more than 65% of the global non-ferrous metals industry's energy consumption. Currently, the industry generally adopts cell control systems based on empirical rules. Process parameters (such as electrode spacing and molecular ratio) mostly rely on manual setting or static model adjustment, with adjustment cycles as long as 2-3 hours. This results in an annual energy waste of approximately 50,000 kWh per cell, leading to a bottleneck where energy efficiency and carbon efficiency are disconnected. Existing technologies focus on single-objective optimization (such as reducing power consumption) and ignore dynamic carbon footprint tracing. Especially in areas with a high proportion of thermal power, where electricity accounts for over 60% of carbon emissions, there is a lack of a mechanism for controlling and regulating process parameters. Existing technologies lack data-driven capabilities. For example, traditional control relies on a few parameters such as cell voltage and temperature, and has weak ability to integrate and analyze multi-dimensional data such as anode current distribution and AlF3 concentration fluctuations, making it difficult to identify "high energy consumption-high emission" conditions in real time (such as local overheating leading to accelerated anode consumption). Existing technologies have a lagging dynamic response. For example, when faced with disturbances such as fluctuations in the thermal balance of the electrolytic cell and frequent anode effects, existing methods have long adjustment cycles (usually on the order of hours), resulting in cumulative energy waste and redundant carbon emissions.

[0003] Although recent studies have attempted to introduce online monitoring, their models are limited to single-dimensional optimization of current efficiency and have not constructed a coupled energy-carbon flow analysis framework. Other technologies, while proposing energy consumption prediction algorithms, have not embedded closed-loop control loops, thus failing to achieve dynamic self-correction of process parameters. Under the "dual carbon" objective, there is an urgent need for an integrated control system that combines multi-source data sensing, energy-carbon synergistic evaluation, and real-time optimization to overcome the core technological bottlenecks in green aluminum electrolysis production. Summary of the Invention

[0004] To address the aforementioned technical problems, this invention provides a system for monitoring, evaluating, and optimizing energy flow and carbon flow in aluminum electrolysis production processes.

[0005] To achieve the above technology, the specific steps are as follows:

[0006] S1. Design the edge layer of the energy flow and carbon flow control system for aluminum electrolysis production process, which includes a sensor network, an edge computing unit, and an actuator; after data acquisition is performed through the sensor network, the edge computing unit performs data processing on the acquired data; the actuator is used for local control.

[0007] Sensor networks are used to perform data acquisition operations, and the network includes: electrical parameter sensors, thermal parameter sensors, chemical parameter sensors, and carbon flow sensors;

[0008] The edge computing unit is used to perform data processing operations on the data collected by the sensor network, including wavelet denoising, Kalman filtering, and anomaly detection.

[0009] The actuators are used for local control, including: regulating the anode lift and the AlF3 addition device;

[0010] The electrical parameter sensors include: a Hall sensor with an error ≤ ±0.1mV, used to measure the tank voltage within a preset time; a Rogowski coil current transformer, used to measure series currents; and a distributed fiber optic sensor, used to measure the anode current distribution.

[0011] The thermal parameter sensors include: an infrared thermometer for measuring the electrolyte temperature within a preset time; and a thermocouple for measuring the tank shell temperature.

[0012] The chemical parameter sensors include: a LIBS laser spectrometer for measuring AlF3 concentration within a preset time; and a conversion ratio (CR) for calculating the molecular ratio.

[0013] The carbon flow sensor includes: a laser displacement sensor for measuring anode consumption; and non-dispersive infrared (NDIR) for continuous monitoring of gas concentration at a fixed sampling frequency.

[0014] S2. The data collected by the edge layer is input into the energy flow-carbon flow coupling model in the cloud layer through the communication layer to perform analysis operations and obtain dynamic evaluation indicators.

[0015] The steps for performing the analysis are as follows:

[0016] S2.1 Construct an energy flow model that includes the input energy equation, the effective energy equation, and the heat loss equation, including the following steps:

[0017] S2.1.1. Considering the energy consumption of the auxiliary system, the input energy equation is constructed by collecting a series of currents and tank voltages and using a preset rectification efficiency. The expression is as follows:

[0018] Q in =I avg ×V cell ×t×η rect +∑P aux ×t

[0019] In the formula, I avg V is the average value of the series current (kA); cell η is the acquired value of the slot voltage; t is the acquisition time; η is the acquisition value of the slot voltage. rect For rectification efficiency; ∑P aux To reduce the energy consumption of auxiliary systems;

[0020] S2.1.2. Based on the aluminum production and the enthalpy of aluminum liquid formation, construct the effective energy equation, the expression of which is as follows:

[0021] Q Al =m Al ×ΔH Al

[0022] In the formula, m Al Aluminum production (kg / h); ΔH Al The enthalpy of aluminum formation;

[0023] S2.1.3. Based on the input energy equation and the effective energy equation, construct the heat loss equation, as shown below:

[0024] Q loss =Q in -Q Al -Q gas

[0025] In the formula, Q gas To remove heat from the flue gas, calculations are performed using a flue gas flow meter and a temperature sensor.

[0026] S2.2 Construct a carbon flow tracking model that includes the anode carbon consumption equation, carbon emission equation, and electricity carbon footprint equation, including the following steps:

[0027] S2.2.1 Based on current efficiency and anode excess consumption coefficient, an anode carbon consumption equation is constructed using a series of collected currents. The expression is as follows:

[0028]

[0029] In the formula, CE is the current efficiency; α is the anode excess consumption coefficient;

[0030] S2.2.2 Based on the carbon oxidation rate, the emission mass of perfluorocarbons (PFCs), and the global warming potential of PFCs, a carbon emission equation is constructed, as shown in the following expression:

[0031]

[0032] In the formula, η os m is the carbon oxidation rate. PFC The emission quality of perfluorocarbons (PFCs); The ratio of CO2 to C is the molecular weight; GWP is the global warming potential of perfluorocarbons (PFCs).

[0033] S2.2.3. Construct the electricity carbon footprint equation based on the input energy equation, with the following expression:

[0034] E CO2 =Q in ×fgrid

[0035] In the formula, f grid The real-time carbon emission factor of the power grid (kg-CO2 / kWh) is obtained through an API interface;

[0036] S2.3 Construct a dynamic evaluation index that includes the comprehensive electricity consumption equation per ton of aluminum, the carbon emission intensity equation per ton of aluminum, and the current efficiency equation, including the following steps:

[0037] S2.3.1 Construct the comprehensive power consumption equation per ton of aluminum, the expression of which is as follows:

[0038]

[0039] S2.3.2 Construct the carbon emission intensity equation per ton of aluminum, the expression of which is as follows:

[0040] C spec =(t-CO2eq / t-Al)

[0041] In the formula, t-CO2eq is the carbon dioxide equivalent; t-Al is the aluminum production (tons);

[0042] S2.3.3 Construct the current efficiency equation, the expression of which is as follows:

[0043]

[0044] S3. Construct an improved multi-objective genetic optimization algorithm (NSGA-II) in the cloud layer, and construct an optimization objective function based on dynamic evaluation indicators; input the optimization objective function into the improved multi-objective genetic optimization algorithm (NSGA-II) and set decision variables and constraints to obtain the optimal combination of process parameters. The steps are as follows:

[0045] S3.1. Use a pre-trained LSTM model to predict the state of the electrolyzer, reduce the search space of the NSGA-II algorithm, and construct an improved multi-objective genetic optimization algorithm (NSGA-II).

[0046] The specific way to reduce the search space of the NSGA-II algorithm by using a pre-trained LSTM model to predict the state of the electrolyzer is as follows: use a 2-layer LSTM (hidden units = 64), Dropout = 0.2, input: historical cell voltage, temperature, and molecular ratio data (time window = 24 hours), output the prediction of the state of the electrolyzer in the next hour, which is used to reduce the search space of the NSGA-II algorithm.

[0047] S3.2 Construct an optimization objective function based on dynamic evaluation indicators, including minimizing power consumption E. spec Minimize carbon emissions C soecTo maximize current efficiency CE, a fitness function, i.e., an objective function, is constructed using real-number encoding, as shown in the following expression:

[0048]

[0049] In the formula, E base This is the baseline value for power consumption; C base This serves as a benchmark for carbon emissions; CE base w1 represents the current efficiency; w2 represents the weight of power consumption; w3 represents the weight of carbon emissions; w1 represents the weight of current efficiency.

[0050] S3.3 Set decision variables, including polar distance, AlF3 addition rate, anode rise and fall frequency, and chromosome length;

[0051] S3.4 Set constraints, including using the penalty function method to limit electrolyte temperature, molecular ratio, and anode effect coefficient; also includes equipment constraints, namely anode lifting speed and AlF3 silo capacity limits;

[0052] S3.5. Input the optimization objective function into the improved multi-objective genetic optimization algorithm (NSGA-II) to obtain the optimal combination of process parameters;

[0053] The population is initialized using an improved multi-objective genetic optimization algorithm (NSGA-II), and the objective function is input to calculate the objective function value. The Pareto front is selected by non-dominated sorting, and a new population is generated through selection, crossover, and mutation to obtain the initial solution. If the number of iterations is less than the maximum number of iterations, the objective function value is recalculated; if the maximum number of iterations is reached, the optimal solution is output.

[0054] The Pareto front screening method is used to select the optimal combination of process parameters from the non-dominated solution set using the fuzzy membership method.

[0055] S4. Based on the improved multi-objective genetic optimization algorithm (NSGA-II) under the optimal combination of process parameters, the data collected by the edge layer sensor network is analyzed and calculated, and dynamic control, closed-loop feedback and abnormal handling operations are performed through the actuator of the edge layer.

[0056] Dynamic control is achieved using actuators, including anode lifting control and AlF3 addition control;

[0057] The anode lifting control is as follows: the anode group is driven to move vertically by a servo motor (the pole pitch is adjusted to an accuracy of ±0.5mm) to adjust the anode position and ensure the pole pitch is stable. A feedforward-feedback composite control is used to compensate for the pole pitch drift caused by anode consumption.

[0058] AlF3 addition control is as follows: feeding is done through a screw feeder with an error of ≤±1% to work with the vibrator to prevent material blockage. When the molecular ratio deviates from the target value of ±0.1, an emergency feeding program is triggered.

[0059] The closed-loop feedback is as follows: data collection is performed every 5 minutes to generate KPIs (power consumption, carbon emissions, current efficiency); the triggering condition is: KPIs deviate from the threshold (power consumption ±3%, carbon emissions ±5%, current efficiency ±1%); the response time is: parameter correction is completed within 30 seconds (by issuing instructions via the OPC protocol).

[0060] Abnormal handling procedures are as follows: if the electrolyte temperature exceeds the limit three times in a row, the emergency cooling system will be activated and the operator will be notified; when a sudden increase in PFC emissions (>10ppm) is detected, the electrode spacing will be automatically reduced and an alumina coating layer will be injected.

[0061] S5. A comparative and economic analysis of the energy flow and carbon flow control system for aluminum electrolysis production process with edge layer, cloud layer and communication layer is conducted.

[0062] The beneficial effects of this invention are:

[0063] This invention achieves comprehensive carbon tracking by incorporating anode consumption, PFC emissions, and electricity carbon footprint into real-time optimization targets, overcoming the limitations of traditional methods that only focus on power consumption. It enhances the model's generalization ability by integrating mechanistic models and data-driven models (LSTM trend prediction). Real-time control is ensured through an edge layer, while the cloud utilizes historical data to train an adaptive optimization model and update weight coefficients. The temperature constraint range is automatically relaxed based on the aging degree of the electrolyzer, extending the cell's lifespan.

[0064] After applying this system to a 300kA prebaked cell series, the following improvements were observed: Energy efficiency improved: DC power consumption per ton of aluminum decreased by 6.2% (12800→12000kWh), resulting in annual electricity savings exceeding 12 million yuan (electricity price 0.5 yuan / kWh); Carbon emission reduction: Carbon emission intensity per ton of aluminum decreased by 10.9% (12.8→11.4t-CO2 eq), resulting in an annual reduction of 75,000 tons of carbon emissions; Process optimization: Current efficiency increased to 92.8%, and the anode effect coefficient decreased from 0.15 to 0.08 times / cell·day. Attached Figure Description

[0065] Figure 1 This is a flowchart of the steps of the present invention;

[0066] Figure 2 This is a system architecture diagram of the present invention;

[0067] Figure 3 This is a schematic diagram of real-time data acquisition according to the present invention;

[0068] Figure 4This is a flowchart of the dynamic evaluation process of the present invention;

[0069] Figure 5 This is a flowchart of the optimization algorithm of the present invention;

[0070] Figure 6 This is a schematic diagram of the feedback mechanism of the present invention. Detailed Implementation

[0071] The present invention will be further described in detail below with reference to specific embodiments.

[0072] like Figure 1 and Figure 2 As shown, a system for monitoring, evaluating, and optimizing energy flow and carbon flow in an aluminum electrolysis production process includes the following steps:

[0073] S1, such as Figure 3 As shown, an edge layer of an energy flow and carbon flow control system for aluminum electrolysis production is designed, incorporating a sensor network, an edge computing unit, and an actuator. After data acquisition is performed via the sensor network, the edge computing unit performs data processing on the acquired data. The actuator is used for local control.

[0074] Specifically, the sensor network is used to perform data acquisition operations, and the network includes: electrical parameter sensors, thermal parameter sensors, chemical parameter sensors, and carbon flow sensors;

[0075] Specifically, the edge computing unit is used to perform data processing operations on the data collected by the sensor network, including wavelet denoising, Kalman filtering, and anomaly detection.

[0076] The edge computing unit hardware includes: a processor (e.g., CPU: ARM Cortex-A72, 8GB memory) supporting at least Modbus / TCP and OPC UA protocols; at least 512GB of memory (e.g., 512GB NVMe SSD); communication interfaces: 4×1Gbps Ethernet, RS485, and CAN bus; the deployment process includes: installing an explosion-proof cabinet (IP54 protection level) in the control room of the electrolysis workshop; configuring a real-time operating system (e.g., ROS2Galactic); and writing a data acquisition driver (e.g., Python / C++) supporting the Modbus / TCP protocol.

[0077] Specifically, the actuators are used for local control, including: regulating the anode lift and the AlF3 addition device;

[0078] Furthermore, the electrical parameter sensors include: a Hall sensor with an error ≤ ±0.1mV for measuring the tank voltage within a preset time; a Rogowski coil current transformer for measuring series currents; and a distributed fiber optic sensor for measuring the anode current distribution.

[0079] In this embodiment, the Hall sensor meets the requirements of a range of 0-5V and an accuracy of ±0.05% FS (such as Honeywell CSLA2CD). Its installation steps include: cleaning the surface of the cathode busbar of the electrolytic cell and applying conductive grease; fixing the sensor with an insulating ceramic bracket to avoid short circuit with the cell shell; connecting the shielded cable to the edge computing unit with a grounding resistance ≤1Ω.

[0080] The Rogowski coil current transformer has a range of 0-500kA and an accuracy of ±0.5%.

[0081] The distributed fiber optic sensor has a spatial resolution of 5cm and is installed on the anode guide rod. The installation steps are as follows: slotting along the longitudinal direction of the anode guide rod and embedding the optical fiber; fixing the optical fiber with high-temperature resistant adhesive (temperature resistance ≥300℃); connecting the optical signal demodulator to the edge computing unit;

[0082] Furthermore, the thermal parameter sensors include: an infrared thermometer for measuring the electrolyte temperature within a preset time; and a thermocouple for measuring the tank temperature.

[0083] In this embodiment, the infrared thermometer has a range of 800-1000℃ and a response time of 50ms (such as the Fluke 62Max+ infrared thermometer). It scans the surface temperature field of the melt in the tank every 30 seconds. The installation steps are as follows: open a Φ50mm observation hole at the top of the tank shell and install a quartz glass window; adjust the pitch angle of the infrared thermometer (30°~45°) to focus on the surface of the melt; calibrate the emissivity (set ε=0.85, corresponding to the molten electrolyte);

[0084] Thermocouples are installed on the outside of the tank shell. 36 thermocouples (such as type K, with an accuracy of ±1℃) are used to construct a three-dimensional temperature field model.

[0085] Furthermore, the chemical parameter sensors include: a LIBS laser spectrometer for measuring AlF3 concentration within a preset time; and a conversion ratio (CR) for calculating the molecular ratio.

[0086] In this embodiment, a LIBS laser spectrometer is used for online analysis of the electrolyte composition, meeting an accuracy of ±0.01mm (e.g., using a Keyence LK-G5000 laser displacement sensor). Data is updated every 5 minutes. The installation steps are as follows: install a bracket on top of the anode assembly to ensure that the laser irradiates the anode surface perpendicularly; calibrate the initial height (H0), and calculate the real-time consumption ΔH = H0 - H(t); combine this with the anode density (1.55 g / cm³). 3 ) Convert carbon consumption rate;

[0087] The ratio of Al2O3 to Na3AlF6 was dynamically calculated based on LIBS measurements by calculating the conversion ratio (CR).

[0088] Furthermore, the carbon flow sensor includes: a laser displacement sensor for measuring anode consumption; and non-dispersive infrared (NDIR) for continuous monitoring of gas concentration at a fixed sampling frequency.

[0089] In this embodiment, a laser displacement sensor is installed on the top of the anode group with an accuracy of ±0.1mm, which is used to calculate the real-time consumption rate by combining the anode height change and density parameters.

[0090] Non-dispersive infrared (NDIR) continuously monitors CO2 and CF4 (a representative PFC) concentrations at a sampling frequency of 1 Hz.

[0091] S2, such as Figure 4 As shown, the data collected by the edge layer is input into the energy flow-carbon flow coupling model in the cloud layer through the communication layer to perform analysis operations and obtain dynamic evaluation indicators.

[0092] The communication layer protocol is: Industrial Ethernet (Modbus / TCP), 5G wireless network, with requirements: latency <100ms, and data encryption: TLS1.3;

[0093] Dynamic evaluation indicators include: comprehensive power consumption equation per ton of aluminum, carbon emission intensity equation per ton of aluminum, and current efficiency equation;

[0094] The steps to perform the analysis are as follows:

[0095] S2.1 Construct an energy flow model that includes the input energy equation, the effective energy equation, and the heat loss equation, including the following steps:

[0096] S2.1.1. Considering the energy consumption of the auxiliary system, the input energy equation is constructed by collecting a series of currents and tank voltages and using a preset rectification efficiency. The expression is as follows:

[0097] Q in =I avg ×V cell ×t×η rect +∑P aux ×t

[0098] In the formula, I avg V is the average value of the series current (kA); cell η represents the tank voltage acquisition value; t represents the acquisition time, acquired every 5 minutes; η represents the acquisition value. rectThe rectification efficiency is set to 96%. The DC power consumption is obtained by multiplying the series of current and slot voltage data collected by the collection time. Multiplying the DC power consumption by the rectification efficiency yields the total energy actually obtained from the AC grid, which is used to compensate for energy losses during the rectification process. ∑P aux Energy consumption of auxiliary systems, including: power consumption of auxiliary equipment such as fans and shell-breaking machines;

[0099] S2.1.2. Based on the aluminum production and the enthalpy of aluminum liquid formation, construct the effective energy equation, the expression of which is as follows:

[0100] Q Al =m Al ×ΔH Al

[0101] In the formula, m Al Aluminum production (kg / h); ΔH Al The enthalpy of aluminum formation, ΔH Al = 8.2 kWh / kg;

[0102] S2.1.3. Based on the input energy equation and the effective energy equation, construct the heat loss equation, as shown below:

[0103] Q loss =Q in -Q Al -Q gas

[0104] In the formula, Q gas To remove heat from the flue gas, calculations are performed using a flue gas flow meter and a temperature sensor.

[0105] S2.2 Construct a carbon flow tracking model that includes the anode carbon consumption equation, carbon emission equation, and electricity carbon footprint equation, including the following steps:

[0106] S2.2.1 Based on current efficiency and anode excess consumption coefficient, an anode carbon consumption equation is constructed using a series of collected currents. The expression is as follows:

[0107]

[0108] In the formula, CE is the current efficiency; α is the anode excess consumption coefficient, with an empirical value of 0.05-0.08, and is taken as 0.05 in this embodiment;

[0109] S2.2.2 Based on the carbon oxidation rate, the emission mass of perfluorocarbons (PFCs), and the global warming potential of PFCs, a carbon emission equation is constructed, as shown in the following expression:

[0110]

[0111] In the formula, η oxThe carbon oxidation rate is taken as 0.98; m PFC The emission quality of perfluorocarbons (PFCs); The ratio of the molecular weight of CO2 to C; GWP is the global warming potential of perfluorinated carbon (PFC) (CF4 = 6630);

[0112] S2.2.3. Construct the electricity carbon footprint equation based on the input energy equation, with the following expression:

[0113] E CO2 =Q in ×f grid

[0114] In the formula, f grid The real-time carbon emission factor of the power grid (kg-CO2 / kWh) is obtained through an API interface;

[0115] S2.3 Construct a dynamic evaluation index that includes the comprehensive electricity consumption equation per ton of aluminum, the carbon emission intensity equation per ton of aluminum, and the current efficiency equation, including the following steps:

[0116] S2.3.1 Construct the comprehensive power consumption equation per ton of aluminum, the expression of which is as follows:

[0117]

[0118] S2.3.2 Construct the carbon emission intensity equation per ton of aluminum, the expression of which is as follows:

[0119] C spec =(t-CO2eq / t-Al)

[0120] In the formula, t-CO2eq is the carbon dioxide equivalent; t-Al is the aluminum production (tons);

[0121] S2.3.3 Construct the current efficiency equation, the expression of which is as follows:

[0122]

[0123] S3. Construct an improved multi-objective genetic optimization algorithm (NSGA-II) in the cloud layer, and construct an optimization objective function based on dynamic evaluation indicators; input the optimization objective function into the improved multi-objective genetic optimization algorithm (NSGA-II) and set decision variables and constraints to obtain the optimal combination of process parameters. The steps are as follows:

[0124] S3.1. Use a pre-trained LSTM model to predict the state of the electrolyzer, reduce the search space of the NSGA-II algorithm, and construct an improved multi-objective genetic optimization algorithm (NSGA-II).

[0125] The specific way to reduce the search space of the NSGA-II algorithm by using a pre-trained LSTM model to predict the state of the electrolyzer is as follows: use a 2-layer LSTM (hidden units = 64), Dropout = 0.2, input: historical cell voltage, temperature, and molecular ratio data (time window = 24 hours), output the prediction of the state of the electrolyzer in the next hour, which is used to reduce the search space of the NSGA-II algorithm.

[0126] S3.2 Construct an optimization objective function based on dynamic evaluation indicators, including minimizing power consumption E. spec Minimize carbon emissions C spec To maximize current efficiency CE, a fitness function, i.e., an objective function, is constructed using real-number encoding, as shown in the following expression:

[0127]

[0128] In the formula, E base The baseline value for electricity consumption is taken as 12800 kWh / t; C base The baseline value for carbon emissions is taken as 12.8 t-CO2 / t; CE base For current efficiency, take 90%; w1 is the weight of power consumption, take 0.4; w2 is the weight of carbon emissions, take 0.4; w3 is the weight of current efficiency, take 0.2.

[0129] S3.3 Set decision variables, including pole distance (10-50mm), AlF addition rate (0.5-5kg / h), anode rise and fall frequency (0.1-1 times / minute), and chromosome length = 3;

[0130] S3.4 Set constraints, including using the penalty function method to limit electrolyte temperature (920-970℃), molecular ratio (2.0-2.5), and anode effect coefficient (<0.1 times / tank·day); also includes equipment constraints, namely anode lifting speed ≤20mm / min and AlF3 silo capacity limit;

[0131] S3.5. Input the optimization objective function into the improved multi-objective genetic optimization algorithm (NSGA-II) to obtain the optimal combination of process parameters;

[0132] like Figure 5 As shown, the population is initialized using the improved multi-objective genetic optimization algorithm (NSGA-II), and the optimization objective function is input to calculate the value of the optimization objective function. The Pareto front is selected by non-dominated sorting, and a new population is generated by selection, crossover, and mutation to obtain the initial solution. If the number of iterations is less than the maximum number of iterations, the value of the optimization objective function is recalculated. If the maximum number of iterations is reached, the optimal combination of process parameters is output.

[0133] In this embodiment, the maximum number of iterations is 500;

[0134] The Pareto front screening method is used to select the optimal combination of process parameters from the non-dominated solution set using the fuzzy membership method.

[0135] S4, such as Figure 6 As shown, the improved multi-objective genetic optimization algorithm (NSGA-II) based on the optimal combination of process parameters is used to analyze and calculate the data collected by the edge layer sensor network, and to perform dynamic control, closed-loop feedback and anomaly handling operations through the actuators of the edge layer.

[0136] Dynamic control is achieved using actuators, including anode lifting control and AlF3 addition control;

[0137] The anode lifting control is as follows: A servo motor drives the anode assembly to move vertically (adjusting the electrode pitch with an accuracy of ±0.5mm) to adjust the anode position and ensure stable electrode pitch. A feedforward-feedback composite control is used to compensate for electrode pitch drift caused by anode consumption. Its control logic includes: target electrode pitch d... target =40mm, allowable fluctuation ±0.5mm; PID parameter: proportional gain K p =1.2, Integral gain K i =0.05, differential gain K d =0.1; the steps are as follows:

[0138] Receive optimization instruction d target ;

[0139] Calculate the anode displacement Δd = d current -d target ;

[0140] Drive the servo motor (meeting a stepping accuracy of 0.01mm) to complete the position adjustment; feed back the actual pole pitch to the edge calculation unit;

[0141] AlF3 addition control is as follows: Feeding is done via a screw feeder with an error of ≤±1%, to work with the vibrator to prevent material blockage. When the molecular ratio deviates from the target value of ±0.1, an emergency replenishment program is triggered. Its control logic is: target addition rate v target =2kg / h, error ≤±1%; screw feeder speed n (rpm) = 0.1×v target The steps are as follows:

[0142] Receive optimization command v target ;

[0143] Adjust the output frequency of the frequency converter to control the speed of the feeder;

[0144] The actual amount of material fed in is fed back by a weighing sensor.

[0145] If the deviation is greater than 2%, the vibrator unblocking procedure will be triggered.

[0146] The closed-loop feedback is as follows: data collection is performed every 5 minutes to generate KPIs (power consumption, carbon emissions, current efficiency); the triggering condition is: KPIs deviate from the threshold (power consumption ±3%, carbon emissions ±5%, current efficiency ±1%); the response time is: parameter correction is completed within 30 seconds (by issuing instructions via the OPC protocol).

[0147] Abnormal handling procedures are as follows: if the electrolyte temperature exceeds the limit three times consecutively, the emergency cooling system will be activated and the operator will be notified; when a sudden increase in PFC emissions (>10ppm) is detected, the electrode spacing will be automatically reduced and an alumina coating layer will be injected.

[0148] This embodiment includes Case 1 and Case 2 for exception handling.

[0149] Case 1: Electrolyte temperature exceeded limit;

[0150] Phenomenon: The temperature of the A07 bath exceeded 970℃ three times consecutively;

[0151] System response: Automatically reduced electrode spacing by 5mm, increased AlF3 addition by 20%, and the temperature recovered to 950℃ within 30 minutes;

[0152] Case 2: Sudden increase in PFC emissions;

[0153] Phenomenon: CF4 concentration in tank A12 > 15 ppm;

[0154] System response: Injection of an alumina coating (thickness increased by 2cm), reduction of electrode spacing by 3mm, and concentration decreased to 5ppm within 1 hour.

[0155] S5. Comparative and economic analysis of the energy flow and carbon flow control system for aluminum electrolysis production process with edge layer, cloud layer and communication layer.

[0156] After applying this system to a certain 300kA prebaked cell series:

[0157] Energy efficiency improvement: DC power consumption per ton of aluminum is reduced by 6.2% (12800→12000kWh), with annual electricity savings exceeding 12 million yuan (electricity price 0.5 yuan / kWh);

[0158] Carbon emission reduction: Carbon emission intensity per ton of aluminum decreased by 10.9% (12.8→11.4t-COeq), resulting in an annual reduction of 75,000 tons of carbon emissions;

[0159] Process optimization: Current efficiency increased to 92.8%, and the anode effect coefficient decreased from 0.15 to 0.08 times / cell·day.

[0160] In this embodiment, the optimal combination of process parameters is compared with the existing technology. The test object for comparison and analysis is: a series of 300kA prebaked anode electrolytic cells (cell numbers A01-A50) of an electrolytic aluminum enterprise; the test period is: January-March 2023 (complete production cycle); the comparison group is: the traditional PID control group (cell numbers B01-B50).

[0161] The results are shown in Table 1 below:

[0162] Table 1: Comparison of Results

[0163] DC power consumption per ton of aluminum 12050kWh 12800kWh ↓5.9％ Carbon emissions per ton of aluminum <![CDATA[11.5t-CO2eq]]> <![CDATA[12.8t-CO2eq]]> ↓10.2％ Current efficiency 92.5％ 90.1％ ↑2.4 percentage points Anode effect coefficient 0.08 times / slot / day 0.15 times / slot / day ↓46.7％

[0164] As shown in the table above, the DC power consumption per ton of aluminum in this invention is reduced by 5.9% compared to the traditional PID, the carbon emission per ton of aluminum is reduced by 10.2% compared to the traditional PID, the current efficiency is increased by 2.4% compared to the traditional PID, and the anode effect coefficient is reduced by 46.7% compared to the traditional PID.

[0165] In addition, an economic analysis of the present invention is provided, including investment costs and annual revenue;

[0166] Investment costs are shown in Table 2;

[0167] Table 2: Investment Costs

[0168] Sensor Networks 120 Edge computing unit 80 Software System 50 Installation and debugging 30 total 280

[0169] Annual returns are shown in Table 3;

[0170] Table 3: Annual Income

[0171] Electricity Saving Benefits 1280 (Energy saving 25.6 million kWh × 0.5 yuan / kWh) Carbon trading revenue 750 (emission reduction of 75,000 tons × 100 yuan / ton) Anode savings 300 (Anode consumption reduced by 8%) total 2330

[0172] Investment payback period: 280 / 2330≈0.12 years (about 1.5 months). Sensitivity analysis shows that when the electricity price fluctuates by ±20%, the payback period varies from 1.2 to 1.8 months.

[0173] Although the present invention has been shown and described through embodiments, those skilled in the art will understand that various modifications, alterations, or substitutions can be made to these embodiments without departing from the principles and spirit of the invention. Therefore, the scope of protection of the present invention should be defined by the appended claims and their equivalents.

Claims

1. A system for monitoring, evaluating, and optimizing energy flow and carbon flow in an aluminum electrolysis production process, characterized in that, Includes the following steps: S1. Design the edge layer of the energy flow and carbon flow control system for aluminum electrolysis production process, which includes a sensor network, an edge computing unit, and an actuator; after data acquisition is performed through the sensor network, the edge computing unit performs data processing on the acquired data; the actuator is used for local control. S2. The data collected by the edge layer is input into the energy flow-carbon flow coupling model in the cloud layer through the communication layer to perform analysis operations and obtain dynamic evaluation indicators. The analysis operation is as follows: by constructing an energy flow model including an input energy equation, an effective energy equation, and a heat loss equation, and by constructing a carbon flow tracking model including an anode carbon consumption equation, a carbon emission equation, and an electricity carbon footprint equation, dynamic evaluation indicators are obtained. The dynamic evaluation indicators include: the comprehensive power consumption equation per ton of aluminum, the carbon emission intensity equation per ton of aluminum, and the current efficiency equation. S3. Construct an improved multi-objective genetic optimization algorithm in the cloud layer, and construct an optimization objective function based on dynamic evaluation indicators; input the optimization objective function into the improved multi-objective genetic optimization algorithm and set decision variables and constraints to obtain the optimal combination of process parameters; The specific steps for S3 are as follows: S3.

1. Use a pre-trained LSTM model to predict the state of the electrolyzer, reduce the search space of the multi-objective genetic optimization algorithm, and construct an improved multi-objective genetic optimization algorithm. The specific way to reduce the search space of the NSGA-II algorithm by using a pre-trained LSTM model to predict the state of the electrolyzer is as follows: use a 2-layer LSTM with Dropout=0.2, input: historical cell voltage, temperature, and molecular ratio data, output the predicted state of the electrolyzer for the next 1 hour, which is used to reduce the search space of the NSGA-II algorithm. S3.2 Construct an optimization objective function based on dynamic evaluation indicators, including minimizing power consumption. Minimize carbon emissions and maximizing current efficiency The fitness function, i.e., the objective function, is constructed using real-number encoding, and its expression is as follows: In the formula, This serves as the baseline value for power consumption; This serves as a benchmark for carbon emissions; For current efficiency; Weighting for power consumption; Weighting for carbon emissions; Weighted by current efficiency; S3.3, Set decision variables, including extreme moments, Addition rate, anode rise / fall frequency, chromosome length; S3.

4. Set constraints, including using the penalty function method to limit electrolyte temperature, molecular ratio, and anode effect coefficient; also including equipment constraints, namely anode rise and fall rates and... Silo capacity limitations; S3.

5. Input the optimization objective function into the improved multi-objective genetic optimization algorithm to obtain the optimal combination of process parameters; S4. Based on the improved multi-objective genetic optimization algorithm under the optimal process parameter combination, the data collected by the edge layer sensor network is analyzed and calculated, and the dynamic control, closed-loop feedback and abnormal handling operations are performed through the edge layer actuator to complete the real-time control of energy flow and carbon flow in the aluminum electrolysis production process. Dynamic control is achieved using actuators, including: anode lifting control and Add control; The anode lifting control is as follows: the anode group is driven to move vertically by a servo motor to adjust the anode position and ensure the electrode pitch is stable. A feedforward-feedback composite control is adopted to compensate for the electrode pitch drift caused by anode consumption. The added control is as follows: feeding is done through a screw feeder with an error of ≤±1% to work with the vibrator to prevent material blockage. When the molecular ratio deviates from the target value of ±0.1, an emergency feeding program is triggered. The closed-loop feedback is as follows: data collection is performed every 5 minutes to generate KPIs; the trigger condition is that the KPI deviates from the threshold; the response time is that parameter correction is completed within 30 seconds. Abnormal handling procedures are as follows: If the electrolyte temperature exceeds the limit three times consecutively, the emergency cooling system will be activated and the operator will be notified; when a sudden increase in PFC emissions is detected, the electrode spacing will be automatically reduced and an alumina coating layer will be injected. S5. A comparative and economic analysis of the energy flow and carbon flow control system for aluminum electrolysis production process with edge layer, cloud layer and communication layer is conducted.

2. The energy flow and carbon flow monitoring, evaluation, and optimization system for an aluminum electrolysis production process according to claim 1, characterized in that, The steps for performing the analysis operation, which involves inputting data collected from the edge layer into the energy flow-carbon flow coupling model in the cloud layer through the communication layer to obtain dynamic evaluation indicators, are as follows: S2.1 Construct an energy flow model that includes the input energy equation, the effective energy equation, and the heat loss equation, including the following steps: S2.1.

1. Considering the energy consumption of the auxiliary system, the input energy equation is constructed by collecting a series of currents and tank voltages and using a preset rectification efficiency. The expression is as follows: In the formula, This is the average value of the series currents; This is the collected value of the slot voltage; For the time of data collection; For rectification efficiency; To reduce the energy consumption of auxiliary systems; S2.1.

2. Based on the aluminum production and the enthalpy of aluminum liquid formation, construct the effective energy equation, the expression of which is as follows: In the formula, For aluminum production; The enthalpy of aluminum formation; S2.1.

3. Based on the input energy equation and the effective energy equation, construct the heat loss equation, as shown below: In the formula, The heat is carried away by the flue gas; S2.2 Construct a carbon flow tracking model that includes the anode carbon consumption equation, carbon emission equation, and electricity carbon footprint equation, including the following steps: S2.2.1 Based on current efficiency and anode excess consumption coefficient, an anode carbon consumption equation is constructed using a series of collected currents. The expression is as follows: In the formula, α is the current efficiency; α is the anode excess consumption coefficient; S2.2.2 Based on the carbon oxidation rate, the emission mass of perfluorinated carbon (PFC), and the global warming potential of PFC, a carbon emission equation is constructed, as shown in the following expression: In the formula, Carbon oxidation rate; The emission quality of perfluorocarbons; for The molecular weight ratio of C; The global warming potential value for perfluorinated carbon; S2.2.

3. Construct the electricity carbon footprint equation based on the input energy equation, with the following expression: In the formula, Real-time carbon emission factor for the power grid; S2.3 Construct a dynamic evaluation index that includes the comprehensive electricity consumption equation per ton of aluminum, the carbon emission intensity equation per ton of aluminum, and the current efficiency equation, including the following steps: S2.3.1 Construct the comprehensive power consumption equation per ton of aluminum, the expression of which is as follows: S2.3.2 Construct the carbon emission intensity equation per ton of aluminum, the expression of which is as follows: In the formula, It is the carbon dioxide equivalent; Aluminum production (tons); S2.3.3 Construct the current efficiency equation, the expression of which is as follows: 。

Images