A method and system for intelligent control of electrolysis temperature field based on infrared thermal imaging and time-series prediction model
The intelligent control method for electrolysis temperature field by using infrared thermal imaging and time-series prediction model solves the problem of untimely electrolysis temperature control, realizes timely control of electrolysis temperature, and improves electrolysis efficiency.
Patent Information
- Application Number
- CN202611304730.X
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-08-26
- Publication Date
- 2026-11-17
AI Technical Summary
The existing electrolysis temperature control system does not adopt intelligent predictive control, resulting in untimely electrolysis temperature control and causing losses.
An intelligent control method for electrolysis temperature field based on infrared thermal imaging and time-series prediction model is adopted. By acquiring discrete temperature data of the three-dimensional temperature measurement matrix, a three-dimensional temperature field and temperature time-series prediction model of the electrolytic cell are constructed, constraints are set, and the optimal actuator coordination scheme is generated.
This enables timely control of electrolysis temperature, reduces the risk of untimely processing, and improves electrolysis efficiency.
Smart Images

Figure CN123108664A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of electrolysis control technology, and more specifically, to an intelligent control method and system for electrolysis temperature field based on infrared thermal imaging and time-series prediction models. Background Technology
[0002] Currently, aluminum electrolysis cells are equipped with high-temperature armored thermocouples and platinum resistance distributed sensor arrays on the sidewalls, cathode steel bars, and cell bottom, with dozens of measuring points deployed in a single cell to collect the temperature of the cell shell, lining, electrolyte, and furnace sides in milliseconds; water electrolysis cells have corrosion-resistant PT100 temperature sensors integrated into the electrode plates and circulation pipelines to collect the temperature of a single pile or single pool zone in real time; copper electrolysis cells have immersion temperature probes deployed in the anode plates and electrolyte to collect the temperature gradient across the entire electrolyte range.
[0003] However, current systems do not employ intelligent predictive control, which may lead to untimely control of electrolysis temperature and result in losses. Summary of the Invention
[0004] The purpose of this invention is to provide a method and system for intelligent control of electrolysis temperature field based on infrared thermal imaging and time-series prediction model, so as to solve the above-mentioned problems in the prior art.
[0005] This invention is achieved through the following technical solution: Firstly, a method for intelligent control of the electrolysis temperature field based on infrared thermal imaging and a time-series prediction model includes: The discrete temperature data of the three-dimensional temperature measurement matrix is obtained, the infrared image of the molten salt surface is obtained and divided into N×N grid regions, and the emissivity of each grid region is corrected. A three-dimensional temperature field of an electrolytic cell is constructed, and a three-dimensional temperature field matrix is obtained from the three-dimensional temperature field of the electrolytic cell based on discrete temperature data. A temperature time series prediction model is constructed. Based on the three-dimensional temperature field matrix, the prediction results are output through the temperature time series prediction model. Several first constraints are set. After the prediction results are corrected based on the constraints, the target prediction result is output. Based on the target prediction results, generate several actuator coordination schemes, construct an objective function and several second constraints, and find the optimal actuator coordination scheme among the several actuator coordination schemes based on the objective function and several second constraints.
[0006] Preferably, the correction of the emissivity for each grid region includes:
[0007] In the formula, For the first Dynamic emissivity of the zone, As the reference emissivity, This represents the real-time deviation between the thermocouple-measured temperature and the infrared temperature measurement within this zone. This is the temperature deviation correction factor for this zone. This is the correction factor for the rate of temperature change in this zone.
[0008] Preferably, the construction of the three-dimensional temperature field of the electrolyzer includes:
[0009] In the formula, The value represents the temperature at the (x, y, z) coordinates of the electrolytic cell. For the three-dimensional coordinates of the electrolytic cell, Let be the spatial coordinates of the i-th actual sensor (or thermocouple). These are the interpolation weighting coefficients. For Kriging trend terms, These are radial basis functions.
[0010] Preferably, the temperature time series prediction model includes a TCN network structure, comprising:
[0011] In the formula, The output of the TCN network at time s. The kernel size is [size]. These are the weighting coefficients of the convolution kernel (filter) at offset i. for, For expansion rate, This is the index offset of the convolution kernel.
[0012] Preferably, the first constraint includes an energy conservation constraint and a temperature change rate constraint, wherein the energy conservation constraint includes:
[0013] In the formula, The specific heat capacity of molten salt, ρ For density, The volume of the molten salt. For Joule heat input, For heat dissipation loss, It is the heat of reaction. To absorb heat during feeding, The temperature of the molten salt. For time, The rate of temperature change; The temperature change rate constraint includes:
[0014] In the formula, and Preset parameters, and Less than B, In minutes.
[0015] Preferred options also include: If the predicted rate of temperature change deviates from the calculated value by the physical equation by more than a threshold, then a physical constraint correction is triggered:
[0016] In the formula, For the final rate of temperature change, The rate of temperature change predicted by the neural network. For adaptive weights, This is the rate of temperature change calculated based on physical equations.
[0017] Preferably, the objective function for construction includes:
[0018] In the formula, To find a set of control sequences that minimizes the objective function J within a given range, The control input calculated at time k and at the current time t. For the future number calculated at time k Step control input, To predict the time domain, Let be the predicted molten salt temperature at time k for the i-th future step. Set the temperature setpoint (target value). This is the weighting matrix for temperature deviation. For the control increment calculated at time k in the future j-th step, The weight matrix for controlling the increment, To control the loop variable (index) in the time domain. This is the index for the current time.
[0019] Secondly, the present invention also provides an intelligent control system for the electrolysis temperature field based on infrared thermal imaging and a time-series prediction model, used to execute the above-mentioned intelligent control method for the electrolysis temperature field based on infrared thermal imaging and a time-series prediction model, including: The data processing module is configured to acquire discrete temperature data from a three-dimensional temperature measurement matrix, acquire an infrared image of the molten salt surface and divide it into N×N grid regions, and correct the emissivity of each grid region; construct a three-dimensional temperature field of the electrolytic cell, and obtain a three-dimensional temperature field matrix based on the discrete temperature data; construct a temperature time series prediction model, output prediction results based on the three-dimensional temperature field matrix, set several first constraints, and output the target prediction result after correcting the prediction results based on the constraints. The command output module is configured to generate several executor coordination schemes based on the target prediction results, construct an objective function and several second constraints, and find the optimal executor coordination scheme among the several executor coordination schemes based on the objective function and several second constraints.
[0020] Thirdly, the present invention also provides a method for obtaining a three-dimensional temperature measurement matrix, comprising: Infrared thermal imagers were deployed at several different heights above the electrolytic cell to observe the molten salt surface temperature field from different perspectives. Several thermocouples are arranged in blind holes inside the tungsten cathode to obtain temperature gradient data in the vertical direction of the molten salt; Several temperature measuring points are arranged radially on the outer surface of the bottom of the electrolytic cell, and the temperature distribution at the bottom of the cell is obtained by using patch thermocouples.
[0021] Preferred options also include: A liftable standard blackbody radiation source is installed inside the electrolytic cell as a dynamic temperature reference. A standard blackbody radiation source is installed on the side wall of the electrolytic cell, and vertical lifting is achieved through a screw mechanism. The thermal imager is dynamically calibrated using the standard temperature of a blackbody radiation source as an absolute reference.
[0022] The technical solution of the present invention has at least the following advantages and beneficial effects: The method provided by this invention mainly includes correcting the emissivity of each grid region; constructing a three-dimensional temperature field of the electrolytic cell, obtaining a three-dimensional temperature field matrix based on discrete temperature data; constructing a temperature time-series prediction model, outputting prediction results based on the three-dimensional temperature field matrix, setting several first constraints, correcting the prediction results based on the constraints, and outputting the target prediction result. Through this method, the optimal execution plan is obtained after filtering the predictions of the current situation based on the temperature time-series prediction model, and then distributed to the corresponding execution mechanism. This allows for timely processing of the current situation, reducing the risk of untimely processing. Attached Figure Description
[0023] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the embodiments will be briefly introduced below. It should be understood that the following drawings only show some embodiments of the present invention and should not be regarded as a limitation on the scope. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.
[0024] Figure 1 This is a schematic diagram of the process of the present invention; Figure 2This is a schematic diagram of the system structure of the present invention. Detailed Implementation
[0025] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. The components of the embodiments of the present invention described and shown in the accompanying drawings can generally be arranged and designed in various different configurations.
[0026] The terms "first," "second," etc., used in the specification, claims, and accompanying drawings of this application are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. The naming or numbering of steps in this application does not imply that the steps in the method flow must be executed in the chronological / logical order indicated by the naming or numbering. The execution order of named or numbered process steps can be changed according to the desired technical objective, as long as the same or similar technical effect is achieved.
[0027] Furthermore, the connection, coupling, or communication in this application can be a direct connection, coupling, or communication between related objects, or an indirect connection, coupling, or communication through other devices. Moreover, the connection, coupling, or communication between objects can be electrical or other similar forms, and this application does not impose any limitations on these. Independently described modules or sub-modules may or may not be physically separated; they may be implemented in software or hardware, and some modules or sub-modules may be implemented in software, with the processor calling the software to implement the function of these modules or sub-modules, while other modules or sub-modules may be implemented in hardware, such as through hardware circuits. Furthermore, some or all of the modules can be selected to achieve the purpose of this application's solution according to actual needs.
[0028] Please refer to Figures 1-2 A method for intelligent control of electrolysis temperature field based on infrared thermal imaging and time-series prediction model, comprising: S1: Obtain discrete temperature data from the three-dimensional temperature measurement matrix, acquire infrared images of the molten salt surface and divide them into N×N grid regions, and correct the emissivity of each grid region; N=5, a total of 25 partitions.
[0029] The emissivity correction for each grid region includes:
[0030] In the formula, For the first Dynamic emissivity of the zone, The reference emissivity is determined by a blackbody reference. This represents the real-time deviation between the thermocouple-measured temperature and the infrared temperature measurement within this zone. This is the temperature deviation correction factor for this zone (obtained through regression of historical data). This is the correction factor for the rate of temperature change in this zone.
[0031] S2: Construct the three-dimensional temperature field of the electrolytic cell, and obtain the three-dimensional temperature field matrix based on discrete temperature data through the three-dimensional temperature field of the electrolytic cell; Constructing the three-dimensional temperature field of the electrolyzer includes:
[0032] In the formula, The value represents the temperature at the (x, y, z) coordinates of the electrolytic cell. For the three-dimensional coordinates of the electrolytic cell, Let be the spatial coordinates of the i-th actual sensor (or thermocouple). These are the interpolation weighting coefficients. For Kriging trend terms, These are radial basis functions.
[0033] This invention additionally extracts the following spatial features of the temperature field: Temperature gradient field: It is used to determine the convection intensity of molten salt.
[0034] Temperature field uniformity index: This quantifies the uniformity of temperature distribution.
[0035] Hotspot or cold spot identification: Detects temperatures that deviate from the global mean by more than [a certain value]. The area marked as abnormal is an area.
[0036] The rate of change of temperature field over time: To determine the trend of temperature changes.
[0037] S4: Construct a temperature time series prediction model, output prediction results based on the three-dimensional temperature field matrix through the temperature time series prediction model, set several first constraints, and output the target prediction result after correcting the prediction results based on the constraints. This invention replaces LSTM with Attention-TCN (Temporal Convolutional Network + Attention Mechanism) and introduces physical constraints to form a hybrid prediction model that combines data-driven and physical constraints.
[0038] Overall model architecture: Input layer (multidimensional time series data), data preprocessing and normalization, TCN dilated convolutional layer (3 layers, dilation rate d=[1,2,4]), multi-head self-attention layer, physical constraint layer (residual correction based on thermodynamic equation), output layer (temperature prediction value and prediction confidence interval).
[0039] The temperature time series prediction model is constructed using a TCN network structure and dilated causal convolution to ensure that no future information is leaked. This includes:
[0040] In the formula, The output of the TCN network at time s. The kernel size is [size]. These are the weighting coefficients of the convolution kernel (filter) at offset i. for, For expansion rate, This is the index offset of the convolution kernel.
[0041] Multi-head self-attention mechanism: A multi-head self-attention layer (8 heads) is connected after the TCN output to automatically identify the importance weight of each input feature (current, voltage, feed rate, temperature of each zone, etc.) to the prediction target:
[0042] In the formula, This is a vector used to query other location information. For use by Matching vectors, Vectors used to provide actual information The length of the feature encoding after linear transformation of the data at each time step. The self-attention mechanism can dynamically identify which sensor data are more important under different operating conditions. For example, under high current conditions, the attention weight of the current feature automatically increases; at the feeding moment, the weight of the feeding amount feature automatically increases.
[0043] A physical constraint layer is added to the output of the neural network, and the prediction results are constrained and corrected using the thermodynamic equations of rare earth molten salt electrolysis. The first constraint includes energy conservation constraint and temperature change rate constraint, wherein the energy conservation constraint includes:
[0044] In the formula, The specific heat capacity of molten salt, ρ For density, The volume of the molten salt. For Joule heat input, For heat dissipation loss, It is the heat of reaction. To absorb heat during feeding, The temperature of the molten salt. For time, The rate of temperature change; The temperature change rate constraint includes:
[0045] In the formula, and Preset parameters, and Less than B means that the temperature change per minute cannot exceed the range of -5℃ to +3℃.
[0046] If the predicted rate of temperature change deviates from the calculated value by the physical equation by more than a threshold, then a physical constraint correction is triggered:
[0047] In the formula, For the final rate of temperature change, The rate of temperature change predicted by the neural network. For adaptive weights, This is the rate of temperature change calculated based on physical equations.
[0048] Predicted output: 1) Temperature point prediction values (future 5 minutes, 10 minutes, 15 minutes).
[0049] 2) Predict confidence interval (95% confidence level).
[0050] 3) Temperature field evolution trend (three-dimensional temperature field prediction for the next 5 minutes).
[0051] S4: Generate several actuator coordination schemes based on the target prediction results, construct an objective function and several second constraints, and find the optimal actuator coordination scheme among the several actuator coordination schemes based on the objective function and several second constraints.
[0052] The traditional PID or fuzzy control is replaced by Model Predictive Control (MPC), and the feedforward-feedback composite control is achieved using the predictive model from step 3. The MPC optimization problem involves constructing the objective function as follows:
[0053] In the formula, To find a set of control sequences that minimizes the objective function J within a given range, The control input calculated at time k and at the current time t. For the future number calculated at time k Step control input, To predict the time domain, Let be the predicted molten salt temperature at time k for the i-th future step. Set the temperature value. This is the weighting matrix for temperature deviation. For the control increment calculated at time k in the future j-th step, The weight matrix for controlling the increment, To control the loop variable in the time domain, This is the index for the current time.
[0054] Second constraint: 1)
[0055] 2) (Current regulation rate limit) 3) (Feeding rate adjustment limit) 4) (Limit on single adjustment of interpole spacing) in (Prediction time domain, corresponding to 15 minutes) (Control time domain), Q and R are weight matrices.
[0056] Collaboration rules: 1) Fast response (second-level temperature fluctuation): prioritize current adjustment (fastest thermal effect).
[0057] 2) Mid-term adjustment (minute-level temperature drift): Adjust the feeding rate (adjust the temperature by absorbing heat during feeding).
[0058] 3) Slow adjustment (hourly temperature trend): Adjust the electrode spacing (change the Joule heat distribution).
[0059] Based on the temperature field prediction results from step S3, adjustment is initiated before a temperature deviation occurs: 1) If it is predicted that the temperature will rise by more than 5°C in the next 5 minutes, reduce the current setting value in advance.
[0060] 2) If it is predicted that the temperature will drop by more than 5°C in the next 10 minutes, increase the feeding rate in advance.
[0061] The measured temperature of the thermocouple is compared with the predicted temperature of the model every 30 seconds, and the prediction error is calculated. The error is fed back to the MPC optimizer to correct the initial state of the next prediction, thus forming a closed-loop control.
[0062] In one instance, regarding the generation and issuance of executor instructions: (1) Current adjustment command The industrial control computer sends the current setting value to the rectifier power supply via the Modbus TCP protocol: 1) Single adjustment range: ≤500A.
[0063] 2) Adjustment cycle: once every 30 seconds.
[0064] 3) Safety protection: If I_set exceeds 90% of the rated current, it must be confirmed by the operator.
[0065] (2) Feeding rate adjustment command Adjust the speed of the feeding screw motor using a frequency converter: 1) Single adjustment range: ≤0.2 kg / min.
[0066] 2) Adjustment cycle: once every 60 seconds.
[0067] 3) Safety protection: The lower limit of the feeding rate is 0.1 kg / min (to prevent material interruption), and the upper limit is 0.5 kg / min (to prevent material caking).
[0068] (3) Pole spacing adjustment command The anode lifting is adjusted by a stepper motor-driven lead screw mechanism: Single adjustment range: ≤5mm.
[0069] Adjustment cycle: once every 300 seconds.
[0070] Safety protection: Lower limit of electrode spacing is 30mm (to prevent short circuit), and upper limit is 100mm (to prevent overvoltage).
[0071] Sensor fault self-diagnosis and fault tolerance: If a thermocouple signal is interrupted or exceeds its range, the system will automatically switch to fusing data from adjacent thermocouples and infrared thermal imaging data to estimate the temperature at that point.
[0072] If the infrared thermal imager lens becomes contaminated, causing the temperature measurement deviation to exceed the threshold, it will automatically trigger compressed gas purging and cleaning. During the cleaning process, it will switch to independent temperature control of the thermocouple array.
[0073] Actuator failure protection: If the rectifier power supply communication is interrupted, it will automatically switch to the backup analog control channel (4–20mA signal).
[0074] If the feeder becomes clogged, the current setting will be automatically reduced (to reduce Joule heat input and prevent overheating), and a clog alarm will be issued at the same time.
[0075] Emergency control under extreme conditions: Emergency control is triggered when the predicted temperature will exceed 1050℃ or fall below 940℃ within 30 seconds: Over-temperature emergency: Immediately cut off the heating current (reduce to open-circuit voltage) and simultaneously increase the feeding rate to the upper limit (use the heat absorption of the feeding to quickly cool down).
[0076] Emergency response to low temperature: Immediately increase the current to 90% of the rated value, and at the same time stop feeding (to reduce heat absorption by cold material).
[0077] Complete control process: Data acquisition: The 3D temperature measurement matrix acquires temperature data in real time; Temperature field reconstruction: zoned emissivity correction and three-dimensional temperature field interpolation; Temperature prediction: Attention-TCN + physically constrained neural network prediction; Control decision: MPC solves for the optimal control sequence; Command issuance: Three actuators coordinate execution; Feedback correction: Measured temperature vs. predicted temperature, correcting the model state; Abnormal self-healing: fault diagnosis and fault-tolerant switching.
[0078] The method provided by this invention mainly includes correcting the emissivity of each grid region; constructing a three-dimensional temperature field of the electrolytic cell, obtaining a three-dimensional temperature field matrix based on discrete temperature data; constructing a temperature time-series prediction model, outputting prediction results based on the three-dimensional temperature field matrix, setting several first constraints, correcting the prediction results based on the constraints, and outputting the target prediction result. Through this method, the optimal execution plan is obtained after filtering the predictions of the current situation based on the temperature time-series prediction model, and then distributed to the corresponding execution mechanism. This allows for timely processing of the current situation, reducing the risk of untimely processing.
[0079] The intelligent control system for electrolysis temperature field based on infrared thermal imaging and time-series prediction model is used to execute the aforementioned intelligent control method for electrolysis temperature field based on infrared thermal imaging and time-series prediction model, including: The data processing module is configured to acquire discrete temperature data from a three-dimensional temperature measurement matrix, acquire an infrared image of the molten salt surface and divide it into N×N grid regions, and correct the emissivity of each grid region; construct a three-dimensional temperature field of the electrolytic cell, and obtain a three-dimensional temperature field matrix based on the discrete temperature data; construct a temperature time series prediction model, output prediction results based on the three-dimensional temperature field matrix, set several first constraints, and output the target prediction result after correcting the prediction results based on the constraints. The command output module is configured to generate several executor coordination schemes based on the target prediction results, construct an objective function and several second constraints, and find the optimal executor coordination scheme among the several executor coordination schemes based on the objective function and several second constraints.
[0080] This invention also provides a method for obtaining a three-dimensional temperature measurement matrix, including: Upper layer (molten salt surface layer): Three infrared thermal imagers are deployed at heights of 1.2m, 1.8m, and 2.5m above the electrolytic cell to observe the molten salt surface temperature field from different perspectives. The three thermal imagers are arranged in a 120° ring to eliminate blind spots caused by obstruction from a single perspective.
[0081] Middle layer (inner layer of molten salt): Three type B thermocouples (with graphite sheaths) are arranged in blind holes inside the tungsten cathode (depths located 50mm, 300mm, and 500mm below the molten salt surface, respectively) to obtain temperature gradient data in the vertical direction of the molten salt.
[0082] Lower layer (bottom layer of the cell): Five temperature measuring points are arranged radially on the outer surface of the bottom of the electrolytic cell (one at the center, two at half the radius, and two at the edge). The temperature distribution at the bottom of the cell is obtained by using patch thermocouples.
[0083] Introduction of dynamic temperature reference: A liftable standard blackbody radiation source is installed inside the electrolytic cell as a dynamic temperature reference. A standard blackbody radiation source (emissivity ≥ 0.995) wrapped in high-temperature resistant ceramic is installed on the side wall of the electrolytic cell, and vertical lifting is achieved through a screw mechanism (stroke 0–300 mm).
[0084] Every 15 minutes, the blackbody radiation source automatically rises and falls to three different heights (the surface of the molten salt, the middle of the molten salt, and the bottom of the molten salt), and the infrared thermal imager simultaneously collects the temperature data of the reference object.
[0085] Using the standard temperature of a blackbody radiation source as an absolute reference, the thermal imager is dynamically calibrated to eliminate measurement drift caused by molten salt volatiles deposited on the lens surface.
[0086] Thermal imager protection status self-diagnosis: A semi-transparent, semi-reflective mirror is installed in front of the lens of each infrared thermal imager. After beam splitting, one beam enters the main optical path of the thermal imager, and the other enters the built-in reference blackbody. By comparing the signal difference between the main optical path and the reference optical path, it is determined in real time whether the lens is corroded by fluoride gas or contaminated by molten salt volatiles. When the difference exceeds a threshold, the lens cleaning device (compressed inert gas purging) is automatically triggered.
[0087] Data acquisition parameters: Infrared thermal imager: resolution 640×512 pixels, temperature measurement range 200–1600℃, accuracy ±1.5℃, sampling frequency 2Hz; Thermocouple: Type B, accuracy class I, sampling frequency 10Hz; Blackbody radiation source calibration frequency: once every 15 minutes; Data upload protocol: EtherCAT industrial Ethernet, synchronization clock accuracy ≤1ms.
[0088] Furthermore, the functional units in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.
[0089] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. This computer software product, stored in a storage medium, includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods of the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.
[0090] The above are merely preferred embodiments of the present invention and are not intended to limit the present invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.
Claims
1. A method for intelligent control of electrolysis temperature field based on infrared thermal imaging and time-series prediction model, characterized in that, include: The discrete temperature data of the three-dimensional temperature measurement matrix is obtained, the infrared image of the molten salt surface is obtained and divided into N×N grid regions, and the emissivity of each grid region is corrected. A three-dimensional temperature field of an electrolytic cell is constructed, and a three-dimensional temperature field matrix is obtained from the three-dimensional temperature field of the electrolytic cell based on discrete temperature data. A temperature time series prediction model is constructed. Based on the three-dimensional temperature field matrix, the prediction results are output through the temperature time series prediction model. Several first constraints are set. After the prediction results are corrected based on the constraints, the target prediction result is output. Based on the target prediction results, generate several actuator coordination schemes, construct an objective function and several second constraints, and find the optimal actuator coordination scheme among the several actuator coordination schemes based on the objective function and several second constraints.
2. The intelligent control method for electrolysis temperature field based on infrared thermal imaging and time-series prediction model according to claim 1, characterized in that, The correction of the emissivity for each grid region includes: In the formula, For the first Dynamic emissivity of the zone, As the reference emissivity, This represents the real-time deviation between the thermocouple-measured temperature and the infrared temperature measurement within this zone. This is the temperature deviation correction factor for this zone. This is the correction factor for the rate of temperature change in this zone.
3. The intelligent control method for electrolysis temperature field based on infrared thermal imaging and time-series prediction model according to claim 1, characterized in that, The construction of the three-dimensional temperature field of the electrolyzer includes: In the formula, The value represents the temperature at the (x, y, z) coordinates of the electrolytic cell. For the three-dimensional coordinates of the electrolytic cell, Let i be the spatial coordinates of the i-th actual sensor. These are the interpolation weighting coefficients. For Kriging trend terms, These are radial basis functions.
4. The intelligent control method for electrolysis temperature field based on infrared thermal imaging and time-series prediction model according to claim 1, characterized in that, The constructed temperature time series prediction model includes a TCN network structure, comprising: In the formula, The output of the TCN network at time s. The kernel size is [size]. These are the weights of the convolution kernel at offset i. for, For expansion rate, This is the index offset of the convolution kernel.
5. The intelligent control method for electrolysis temperature field based on infrared thermal imaging and time-series prediction model according to claim 1, characterized in that, The first constraint includes an energy conservation constraint and a temperature change rate constraint, wherein the energy conservation constraint includes: In the formula, The specific heat capacity of molten salt, ρ For density, The volume of the molten salt. For Joule heat input, For heat dissipation loss, The heat of reaction, To absorb heat during feeding, The temperature of the molten salt. For time, The rate of temperature change; The temperature change rate constraint includes: In the formula, and Preset parameters, and Less than B, In minutes.
6. The intelligent control method for electrolysis temperature field based on infrared thermal imaging and time-series prediction model according to claim 5, characterized in that, Also includes: If the predicted rate of temperature change deviates from the calculated value by the physical equation by more than a threshold, then a physical constraint correction is triggered: In the formula, For the final rate of temperature change, The rate of temperature change predicted by the neural network. For adaptive weights, This is the rate of temperature change calculated based on physical equations.
7. The intelligent control method for electrolysis temperature field based on infrared thermal imaging and time-series prediction model according to claim 5, characterized in that, The objective function for construction includes: In the formula, To find a set of control sequences that minimizes the objective function J within a given range, The control input calculated at time k and at the current time t. For the future number calculated at time k Step control input, To predict the time domain, Let be the predicted molten salt temperature at time k for the i-th future step. Set the temperature value. This is the weighting matrix for temperature deviation. For the control increment calculated at time k in the future j-th step, The weight matrix for controlling the increment, To control the loop variable in the time domain, This is the index for the current time.
8. An intelligent control system for electrolysis temperature field based on infrared thermal imaging and time-series prediction model, characterized in that, A method for intelligent control of electrolysis temperature field based on infrared thermal imaging and time-series prediction model as described in any one of claims 1-7, comprising: The data processing module is configured to acquire discrete temperature data from a three-dimensional temperature measurement matrix, acquire an infrared image of the molten salt surface and divide it into N×N grid regions, and correct the emissivity of each grid region; construct a three-dimensional temperature field of the electrolytic cell, and obtain a three-dimensional temperature field matrix based on the discrete temperature data; construct a temperature time series prediction model, output prediction results based on the three-dimensional temperature field matrix, set several first constraints, and output the target prediction result after correcting the prediction results based on the constraints. The command output module is configured to generate several executor coordination schemes based on the target prediction results, construct an objective function and several second constraints, and find the optimal executor coordination scheme among the several executor coordination schemes based on the objective function and several second constraints.
9. A method for obtaining a three-dimensional temperature measurement matrix, characterized in that, A method for intelligent control of electrolysis temperature field based on infrared thermal imaging and time-series prediction model as described in any one of claims 1-7, comprising: Infrared thermal imagers were deployed at several different heights above the electrolytic cell to observe the molten salt surface temperature field from different perspectives. Several thermocouples are arranged in blind holes inside the tungsten cathode to obtain temperature gradient data in the vertical direction of the molten salt; Several temperature measuring points are arranged radially on the outer surface of the bottom of the electrolytic cell, and the temperature distribution at the bottom of the cell is obtained by using patch thermocouples.
10. The method for obtaining a three-dimensional temperature measurement matrix according to claim 9, characterized in that, Also includes: A liftable standard blackbody radiation source is installed inside the electrolytic cell as a dynamic temperature reference. A standard blackbody radiation source is installed on the side wall of the electrolytic cell, and vertical lifting is achieved through a screw mechanism. The thermal imager is dynamically calibrated using the standard temperature of a blackbody radiation source as an absolute reference.
