Event-triggered method and system for controlling alumina concentration in an aluminium reduction cell

Abstract

The application provides an event-triggered alumina concentration control method and system for an aluminum electrolysis cell, and belongs to the field of electrolytic cell discharging control. The method analyzes the dynamic balance process of the alumina concentration of aluminum electrolysis, constructs an alumina concentration dynamic balance model and discretizes it, constructs an alumina dissolution discrete time model, and then converts it into an error model. An alumina concentration controller is constructed, including an execution network, an error model and an evaluation network. The execution network is trained offline. Real-time data of the aluminum electrolysis cell at the current time is collected. An event-triggered mechanism is used to determine whether the real-time data at the current time meets the triggering condition. If it does, the real-time data is input into the controller, and the optimal discharging amount at the next time is output. If it does not, the discharging amount at the current time is output as the optimal discharging amount at the next time. The application realizes the control of the alumina concentration of the electrolytic cell, improves the accuracy and stability of the concentration control, and improves the production efficiency and quality of aluminum electrolysis.

Description

Technical Field

[0001] This invention belongs to the field of electrolytic cell feeding control, specifically relating to an event-triggered method and system for controlling alumina concentration in aluminum electrolytic cells. Background Technology

[0002] Aluminum, as an important industrial metallic material, is smelted through processes including the preparation of alumina and the production of electrolytic aluminum. The quality of aluminum metal is directly determined by the aluminum electrolysis industry. Figure 1 As shown, the electrolytic aluminum production process generally uses cryolite-alumina molten salt, with molten cryolite as the solvent and alumina as the solute. Carbonaceous material is used as the anode, and molten aluminum as the cathode. After a strong direct current is applied, an electrochemical reaction occurs between the anode and cathode in an electrolytic cell at high temperature, thus producing aluminum metal. Achieving stable control of the aluminum electrolysis process is crucial for optimizing the operating state of the electrolytic cell, improving current efficiency, and reducing energy consumption. As one of the core parameters for electrolysis process control, the precise control of alumina concentration directly affects the thermal balance and current efficiency of the electrolytic cell. If the alumina concentration is too low, an anodic effect easily occurs in the electrolytic cell, which not only disrupts the stability of the electrolysis process but also reduces current efficiency. If the alumina concentration is too high, insoluble precipitates easily form at the bottom of the cell, accelerating cathode corrosion, shortening the cell life, and increasing maintenance costs. Achieving precise control of alumina concentration is key to ensuring the efficient, energy-saving, and stable operation of the aluminum electrolysis process, and is of great significance for promoting the sustainable development of the aluminum electrolysis industry. The concentration of alumina depends on the amount of alumina fed into the electrolytic cell, which is essentially controlled by controlling the amount of alumina fed into the cell. Therefore, controlling the concentration of alumina in the electrolytic cell is essentially controlling the amount of alumina fed into the cell.

[0003] In existing technologies, traditional control methods for controlling alumina feed rate include timing control, fuzzy control, and cell resistance control, as well as predictive control methods based on machine learning models. Traditional control methods rely on expert experience and manual operation, which cannot match the multivariate coupling and time-varying characteristics of electrolytic cells. Especially when cell conditions are unstable, the control effect is poor, affecting production efficiency and quality. While neural network-based predictive control methods perform well in simulations, they are limited by strong dependence on training data, insufficient generalization ability, and susceptibility to local optima. Multivariate control strategies combining mechanistic models (such as model predictive control) improve control accuracy, but due to high computational complexity and strong dependence on model accuracy, they are difficult to guarantee real-time performance and robustness in actual industrial scenarios. However, due to the large number of time-varying parameters and unmodeled dynamic characteristics in the aluminum electrolysis process, it is difficult to establish accurate models, and the accuracy and stability of alumina feed rate control in actual aluminum electrolysis industrial processes cannot meet industrial requirements. Summary of the Invention

[0004] To address the aforementioned issues, this invention provides an event-triggered method and system for controlling alumina concentration in an aluminum electrolysis cell. Based on the process characteristics of aluminum electrolysis, an aluminum electrolysis process model describing the dynamic changes of the system is constructed. An ADP feeding controller incorporating a fuzzy neural network is designed. Initial permissible control is obtained through offline training, and the optimal alumina feeding amount is solved online through optimization. Furthermore, an event-triggered mechanism is introduced to control the alumina concentration in the electrolysis cell, improving the accuracy and stability of concentration control and enhancing the production efficiency and quality of aluminum electrolysis.

[0005] To achieve the above objectives, the technical solutions adopted in the embodiments of the present invention are as follows:

[0006] In a first aspect, embodiments of the present invention provide a method for controlling alumina concentration in an aluminum electrolytic cell based on event triggering, the method comprising the following steps:

[0007] Step S1: Analyze the dynamic equilibrium process of alumina concentration in aluminum electrolysis and construct a dynamic equilibrium model of alumina concentration.

[0008] Step S2: Discretize the dynamic equilibrium model of alumina concentration to construct a discrete time model of alumina dissolution.

[0009] Step S3: Convert the discrete-time model of alumina dissolution into an error model;

[0010] Step S4: Construct an alumina concentration controller based on an error model; the controller includes an execution network, an error model, and an evaluation network.

[0011] Step S5: Train the execution network offline;

[0012] Step S6: Collect the real-time data of the aluminum electrolysis cell at the current moment; determine whether the real-time data of the material at the current moment meets the triggering conditions based on the event triggering mechanism; if it meets the conditions, execute step S7; if it does not meet the conditions, output the current feeding amount as the optimal feeding amount for the next moment.

[0013] Step S7: Input the real-time data into the controller and output the optimal feeding amount for the next moment.

[0014] In a preferred embodiment of the present invention, step S1, which involves constructing a dynamic equilibrium model for alumina concentration, includes:

[0015] Step S11: Analyze the dual kinetic characteristics of the alumina dissolution process in aluminum electrolysis and construct an alumina dissolution model;

[0016] Step S12: Based on the alumina dissolution model and Faraday's law of electrolysis, a dynamic equilibrium model of alumina concentration is constructed.

[0017] In a preferred embodiment of the present invention, the alumina dissolution model formula is as follows:

[0018] (1)

[0019] (2)

[0020] In equations (1) and (2), It is the mass fraction of alumina powder at time t. It is the solubility constant of alumina powder; It is the alumina feeding rate at time t. The proportion is the ratio of powder to powder; It is the mass fraction of alumina aggregates. It is the solubility constant of alumina aggregates; It is the electrolysis temperature; It is the reference electrolysis temperature; It is an empirical constant; This represents the mass of the electrolyte melt.

[0021] As a preferred embodiment of the present invention, the dynamic equilibrium model of alumina concentration is as follows:

[0022] (3)

[0023] In equation (3), Let be the alumina concentration at time t; Here is the molar mass of aluminum oxide; For current efficiency; For series current; It is Faraday's constant; The number of electrons transferred.

[0024] In a preferred embodiment of the present invention, when discretizing the dynamic equilibrium model of alumina concentration in step S2, alumina concentration is used as the main controlled variable, and the mass fraction of alumina powder and the mass fraction of alumina agglomerates are used as auxiliary controlled variables. The relationship between the main and auxiliary variables is quantified by controlling the weights of the three variables. The alumina feeding rate is used as the control variable.

[0025] In a preferred embodiment of the present invention, the discrete-time model for alumina dissolution is as follows:

[0026] (4)

[0027] In equation (4), for The state variables related to alumina concentration at time and The utility function state penalty quadratic form is ,in, and ; for Alumina feeding rate at all times; Electrolytic consumption during the alumina dissolution process and ; Represents the dynamic matrix of alumina dissolution. This represents the input matrix for alumina feeding.

[0028] In a preferred embodiment of the present invention, when step S3 is converted into an error model, the reference trajectory of alumina concentration is defined as follows: The trajectory function that satisfies for:

[0029] (5)

[0030] The tracking error is The steady-state control variable at this point is defined as:

[0031] (6)

[0032] The error model is:

[0033] (7)

[0034] In equation (7), This represents the error control amount at time k;

[0035] Define the utility function as follows:

[0036] (8)

[0037] In equation (8), satisfy and ; R is a penalty factor for alumina concentration tracking error and feeding rate;

[0038] Define the performance index function as follows:

[0039] (9)

[0040] The optimal performance index function is expressed as:

[0041] (10)

[0042] Optimal cost function The optimal error control quantity can be obtained by satisfying the HJB equation. for:

[0043] (11).

[0044] In a preferred embodiment of the present invention, the input to the execution network in step S4 is Tracking error at any time The output is Optimal error control quantity at time 1 The input to the error model is Moment and The output is Tracking error at any time ; Evaluate the network input The output is Optimal performance index function at time 1 .

[0045] In a preferred embodiment of the present invention, when determining whether the triggering condition is met in step S6, the event triggering time is defined as a monotonically increasing time sequence. At the triggering time Update the material feeding amount in the control system. The feeding control strategy is saved to the zero-order hold; feeding amount Represented as:

[0046] (12)

[0047] In equation (12), , ;

[0048] At non-event-triggered times Zero-order hold control strategy until the next event is triggered;

[0049] Define the error trigger for material feed update for:

[0050] (13)

[0051] In equation (13), express The tracking error of alumina concentration at any given time. This represents the tracking error at the s-th trigger moment;

[0052] definition This is the error control input for the error system; the optimal feed rate is:

[0053] (14)

[0054] In equation (14), This represents the optimal feed rate at the s-th trigger time;

[0055] The trigger threshold conditions are defined as shown in equations (15) and (16):

[0056] (15)

[0057] (16)

[0058] In equations (15) and (16), and This is the trigger threshold;

[0059] The expression for setting the event trigger threshold is:

[0060] (17)

[0061] (18)

[0062] In equations (17) and (18), It is a constant;

[0063] Online iteration ensures the system's adaptive control: loading the initial allowable control obtained from offline training. If the event triggering condition (15) or (16) is met, then training the evaluation network yields a value function that satisfies the policy evaluation condition, and training the execution network yields a control policy that satisfies the policy improvement condition; if in Internally satisfying the error condition Then update the optimal feed rate. The alumina concentration at the next moment is obtained by inputting it into the aluminum electrolysis process model. Otherwise, repeat the iterative training.

[0064] If the event triggering conditions (15) and (16) are not met, then calculate directly. The alumina concentration at the next moment is obtained by inputting it into the aluminum electrolysis process model. .

[0065] Secondly, embodiments of the present invention also provide an event-triggered alumina concentration control system for aluminum electrolysis cells. The system includes an alumina concentration dynamic equilibrium model construction module, a model discretization module, an error module construction module, a controller construction module, an execution network training module, a trigger condition judgment module, a controller execution module, and a feed rate output module.

[0066] The alumina concentration dynamic equilibrium model construction module is used to analyze the alumina concentration dynamic equilibrium process in aluminum electrolysis and construct an alumina concentration dynamic equilibrium model.

[0067] The model discretization module is used to discretize the alumina concentration dynamic equilibrium model and construct an alumina dissolution discrete time model.

[0068] The error module construction module is used to convert the discrete-time model of alumina dissolution into an error model;

[0069] The controller construction module is used to construct an alumina concentration controller based on an error model; the controller includes an execution network, an error model, and an evaluation network.

[0070] The execution network training module is used to perform offline training on the execution network;

[0071] The trigger condition judgment module is used to collect the real-time data of the aluminum electrolysis cell at the current moment; it judges whether the real-time data of the material at the current moment meets the trigger condition based on the event triggering mechanism; if it does, the controller execution module is started; if it does not meet, the material feeding amount at the current moment is used as the material feeding amount at the next moment, and the material feeding amount output module is started directly.

[0072] The controller execution module is used to input real-time data into the controller to generate the optimal feeding amount for the next moment;

[0073] The material feeding output module is used to output the optimal material feeding amount for the next moment.

[0074] The solutions of the embodiments of the present invention have the following beneficial effects:

[0075] The event-triggered alumina concentration control method and system for aluminum electrolysis cells provided in this invention significantly reduces the computational load while ensuring accurate tracking and control of alumina concentration. The designed triggering conditions only initiate control updates when the system state error exceeds a preset threshold, effectively reducing unnecessary calculations by the controller and significantly improving its applicability in the aluminum electrolysis industrial environment. Simultaneously, the fuzzy neural network structure enhances the adaptive capability to the time-varying and multivariate coupling characteristics of the aluminum electrolysis process, rigorously proving the eventual uniform bounded stability of the closed-loop system, and providing an efficient and reliable control solution for the intelligent upgrading of the aluminum electrolysis industry.

[0076] Of course, implementing any product or method of the present invention does not necessarily require achieving all of the advantages described above at the same time. Attached Figure Description

[0077] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0078] Figure 1 This is a schematic diagram of a single aluminum electrolytic cell structure based on the Hall-Erou method in the prior art;

[0079] Figure 2 This is a schematic diagram of the event-triggered alumina concentration control method for aluminum electrolysis cells according to an embodiment of the present invention.

[0080] Figure 3 This is a flowchart of the event-triggered alumina concentration control method for aluminum electrolysis cells according to an embodiment of the present invention;

[0081] Figure 4 This is a control principle diagram based on event triggering as described in an embodiment of the present invention. Detailed Implementation

[0082] After discovering the aforementioned problems, the inventors of this application conducted a detailed study on existing methods for controlling alumina concentration in aluminum electrolysis processes. The study revealed that the aluminum electrolysis process inherently possesses complex characteristics, including infinite-dimensional states, high nonlinearity, and strong variable coupling. In complex industrial process control, the Adaptive Dynamic Programming (ADP) method, which integrates the advantages of reinforcement learning and neural networks, effectively solves the curse of dimensionality problem inherent in dynamic programming. The core idea of ​​the ADP algorithm is to obtain the optimal solution of the Hamilton-Jacobi-Bellman (HJB) equation through iterative learning, thereby achieving optimal control with low model dependence. However, existing ADP algorithms generally use backpropagation neural networks (BP) to approximate the evaluation and action networks, resulting in controllers with insufficient robustness and weak anti-interference capabilities. In contrast, fuzzy neural networks (FNNs), by integrating the advantages of neural networks and fuzzy theory, can not only effectively solve uncertainties that traditional neural networks struggle to handle, but also exhibit stronger adaptive capabilities when dealing with complex systems. Meanwhile, with the expansion of aluminum electrolysis production scale and the improvement of intelligent level, the computing efficiency and hardware cost of the control system are among the key considerations.

[0083] Therefore, by introducing an event trigger (ET) mechanism, unnecessary control updates of traditional time-triggered ADP can be reduced, and FNN can be used instead of traditional BP neural network to approximate the execution network. This significantly reduces the computational load while ensuring accuracy, providing an efficient and feasible control strategy for the large-scale and intelligent development of the aluminum electrolysis industry.

[0084] It should be noted that the defects in the above-mentioned prior art solutions are all the result of the inventors' practice and careful research. Therefore, the discovery process of the above problems and the solutions proposed by the embodiments of the present invention in the following text should be the inventors' contributions to the present invention.

[0085] 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 a part of the embodiments of the present invention, and not all of them. 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. It should be noted that, without conflict, the embodiments and features in the embodiments of the present invention can also be combined with each other.

[0086] It should be noted that similar reference numerals and letters in the following figures indicate similar items. Therefore, once an item is defined in one figure, it does not need to be further defined and explained in subsequent figures. In the description of the embodiments of the present invention, the terms "first," "second," "third," "fourth," etc., are used only to distinguish descriptions and should not be construed as indicating or implying relative importance. In addition, sometimes a subscript such as W1 may be written in a non-subscript form such as W1, and their meanings are consistent unless the distinction is emphasized.

[0087] It should be understood that the term "and / or" in this article is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A alone, A and B simultaneously, or B alone. A and B can be singular or plural. Additionally, the character " / " in this article generally indicates an "or" relationship between the preceding and following related objects, but it can also represent an "and / or" relationship. Please refer to the context for a more accurate understanding.

[0088] After the above in-depth analysis, the embodiments of the present invention provide an event-triggered method and system for controlling alumina concentration in an aluminum electrolysis cell. Based on the process characteristics of the aluminum electrolysis process, an aluminum electrolysis process model describing the dynamic changes of the system is constructed, and an ADP feeding controller integrating a fuzzy neural network is designed. The initial permissible control is obtained through offline training, and the optimal alumina feeding amount is solved online through optimization. Furthermore, an event-triggered mechanism is introduced, which significantly improves the utilization efficiency of the controller's computing resources while ensuring the stability of the controller. The stability and effectiveness of the controller are verified through simulation experiments.

[0089] The event-triggered alumina concentration control method for aluminum electrolysis cells can be implemented by electronic devices, such as terminals or servers. Figure 2 and Figure 3As shown, the method includes:

[0090] Step S1: Analyze the dynamic equilibrium process of alumina concentration in aluminum electrolysis and construct a dynamic equilibrium model of alumina concentration.

[0091] The aluminum electrolysis process involved in this step generally refers to, for example, the aluminum electrolysis process. Figure 1 The process shown employs cryolite-alumina molten salt electrolysis, and the equipment uses a large prebaked anode aluminum electrolysis cell with a capacity of over 200kA. During electrolysis, the anode and molten aluminum participate in the reaction as the anode and cathode, respectively, and the anode material is generally carbon. The alumina raw material is transported to the electrolysis cell feeding hopper through a high-pressure dense phase conveying system. The cell control system calculates the feeding frequency based on real-time operating conditions, and then the alumina is added to the electrolyte through the feeding mechanism. Under high temperature conditions (such as 935~965℃), a strong direct current is applied to induce an electrochemical reaction between the two electrodes as shown in equation (1):

[0092] (1)

[0093] The construction of the dynamic equilibrium model for alumina concentration includes:

[0094] Step S11 analyzes the dual kinetic characteristics of the alumina dissolution process in aluminum electrolysis, including: rapid dissolution of fine alumina powder in the electrolyte; and gradual dissolution of alumina aggregates at a relatively slow rate. An alumina dissolution model is constructed based on these dual kinetic characteristics in the electrolytic cell, as follows:

[0095] (2)

[0096] (3)

[0097] In equations (2) and (3), It is the mass fraction of alumina powder at time t. It is the solubility constant of alumina powder; It is the alumina feeding rate at time t. The proportion is the ratio of powder to powder; It is the mass fraction of alumina aggregates. It is the solubility constant of alumina aggregates; It is the electrolysis temperature; It is the reference electrolysis temperature; It is an empirical constant; This represents the mass of the electrolyte melt.

[0098] Step S12: Based on the alumina dissolution model and Faraday's law of electrolysis, a dynamic equilibrium model of alumina concentration is constructed. Under high-temperature electrolysis conditions, dissolved alumina undergoes electrochemical reduction at the cathode surface to form metallic aluminum, while oxygen ions are oxidized at the anode to form carbon dioxide. According to Faraday's law of electrolysis, the dynamic equilibrium model of alumina concentration is as follows:

[0099] (4)

[0100] In equation (4), Let be the alumina concentration at time t; Here is the molar mass of aluminum oxide; For current efficiency; For series current; It is Faraday's constant; The number of electrons transferred.

[0101] Step S2: Discretize the dynamic equilibrium model of alumina concentration to construct a discrete time model of alumina dissolution.

[0102] In this step, when discretizing the dynamic equilibrium model of alumina concentration, the Euler method is used for discretization during the alumina concentration control stage, assuming a constant electrode spacing. This adapts to the environment where aluminum electrolysis process control systems typically employ discrete-time sampling and control. In actual industrial processes, both electrode spacing and alumina concentration affect the cell voltage, which is controlled by adjusting both. The electrode spacing adjustment interval is much longer than the alumina concentration adjustment interval; therefore, it can be assumed that the electrode spacing remains constant during alumina concentration adjustment.

[0103] Specifically, alumina concentration is used as the primary controlled variable, while the mass fraction of alumina powder and the mass fraction of alumina agglomerates are used as auxiliary controlled variables. The primary and auxiliary relationships are quantified by controlling the weights of these three variables. The alumina feeding rate is used as a control variable, and the discrete-time model for alumina dissolution is as follows:

[0104] (5)

[0105] In equation (5), for The state variables related to alumina concentration at time and The utility function state penalty quadratic form is ,in, and ; for Alumina feeding rate at all times; Electrolytic consumption during the alumina dissolution process and ; Represents the dynamic matrix of alumina dissolution. The alumina feeding input matrix is ​​represented as follows:

[0106] (6)

[0107] (7)

[0108] In equations (6) and (7), Indicates the sampling interval; and These represent the solubility constants of alumina powder and agglomerates, respectively. Indicates the solubility ratio of the powder; These are empirical constants; and These represent the electrolysis temperature and the reference electrolysis temperature, respectively. This indicates the mass of the electrolyte melt.

[0109] Step S3: Convert the discrete-time model of alumina dissolution into an error model.

[0110] In this step, when converting to an error model, the alumina concentration reference trajectory is defined as follows: The trajectory function that satisfies for:

[0111] (8)

[0112] The tracking error is The steady-state control variable at this point is defined as:

[0113] (9)

[0114] The error model is:

[0115] (10)

[0116] In equation (10), This represents the error control quantity at time k.

[0117] Define the utility function as follows:

[0118] (11)

[0119] In equation (11), satisfy and Q and R are penalty factors for alumina concentration tracking error and feeding rate. This quadratic performance index function not only enables the system state to track the reference trajectory, but also ensures that the system control input remains stable during the process of tracking the reference value.

[0120] Define the performance index function as follows:

[0121] (12)

[0122] According to Bellman's optimality principle, the optimal performance index function can be expressed as:

[0123] (13)

[0124] Optimal cost function The optimal error control quantity can be obtained by satisfying the HJB equation. for:

[0125] (14).

[0126] Step S4: Construct an alumina concentration controller based on an error model; the controller includes an execution network, an error model, and an evaluation network.

[0127] In this step, the input to the execution network is The output is the optimal error control quantity. The input to the error model is and The output is Evaluate the network input. The output is the optimal performance index function. .

[0128] The hidden layers of both the execution and evaluation networks use tanh as the activation function. Considering that FNNs can effectively handle uncertainty and possess stronger interpretability and adaptability, this step uses an FNN instead of a traditional BP neural network to approximate the execution network.

[0129] (1) Execute network

[0130] A fuzzy neural network consists of: an input layer, a fuzzification layer, a rule layer, a normalization layer, a defuzzification layer, and an output layer.

[0131] The input layer will input vectors The data is passed to the blurring layer, which contains n neurons.

[0132] In the fuzzification layer, each neuron represents a fuzzy set, and the membership degree of each input to each fuzzy set is calculated:

[0133] (15)

[0134] In equation (15), ; ; n is the dimension of the input variable; for The fuzzy segmentation number. A Gaussian function is chosen as the membership function, with the specific form shown in equation (20):

[0135] (16)

[0136] In equation (16), It is the mean of the function; This is the function width. This layer contains a total of... One neuron.

[0137] In the rule layer, each neuron represents a fuzzy rule, and the applicability of that fuzzy rule is calculated:

[0138] (17)

[0139] In equation (17), ; And satisfy This layer contains a total of m neurons.

[0140] The normalization layer normalizes the rules to accurately represent their relative importance.

[0141] (18)

[0142] In equation (18), This layer contains a total of m neurons.

[0143] The output layer deblurs the normalized outputs and calculates the total output of all nodes:

[0144] (19)

[0145] In equation (19), ; The number of fuzzy rules; For weights.

[0146] Therefore, the execution network is used to approximate the optimal control strategy, taking the system state error as the input variable, and defining the output of the execution network's feed rate. With target control quantity There are errors :

[0147] (20)

[0148] Objective function for network training It is in the following form:

[0149] (twenty one)

[0150] Update the weights using gradient descent:

[0151] (twenty two)

[0152] In equation (22), This represents the weights at time k when the network training is performed for the jth time.

[0153] (2) Evaluation Network

[0154] The evaluation network is used to approximate the optimal performance index function, taking the system state error as input and defining the output of the evaluation network. With the target performance index function There are errors :

[0155] (twenty three)

[0156] The objective function for evaluating network training is as follows:

[0157] (twenty four)

[0158] Update the weights using gradient descent:

[0159] (25)

[0160] In equation (25), This represents the weights at time k when evaluating the network during the j-th training iteration.

[0161] Step S5: Perform offline training on the execution network.

[0162] In this step, the initial weights of the execution network based on the fuzzy neural network are randomly selected. If the selected weights can meet the described error conditions, the offline training is considered successful; otherwise, the neural network weights are reselected.

[0163] Online iteration requires significantly more computation when initial values ​​are not ideal, and may even fail to guarantee eventual convergence. Therefore, before policy iteration begins, offline training is necessary to obtain initial, permissible control: given the initial action network and evaluation network parameters, the maximum number of iterations... and training error Assume the initial control output value Initial state error The evaluation network is trained according to equations (23) to (25). If in Internal conditions are met Then, the initial evaluation network and execution network parameters are saved, and the following definitions are defined. As the initial permissive control for online iteration.

[0164] Step S6: Collect the real-time data of the aluminum electrolysis cell at the current moment; determine whether the real-time data of the material at the current moment meets the triggering conditions based on the event triggering mechanism; if it meets the conditions, execute step S7; if it does not meet the conditions, output the material feeding amount at the current moment as the material feeding amount at the next moment.

[0165] In this step, such as Figure 4 As shown, a trigger threshold is set to determine whether the triggering condition is met. The event trigger time is defined as a monotonically increasing time sequence. At the triggering time Update the material feeding amount in the control system. And the feeding control strategy is saved to the zero-order hold. Feed amount Represented as:

[0166] (26)

[0167] In equation (26), , At non-event-triggered times Zero-order hold control strategy Until the next event is triggered.

[0168] Define the error trigger for material feed update for:

[0169] (27)

[0170] In equation (27), express The tracking error of alumina concentration at any given time. This represents the tracking error at the s-th trigger moment. (Definition) This is the error control input for the error system. The optimal feed rate is:

[0171] (28)

[0172] In equation (28), This represents the optimal feed rate at the s-th trigger time.

[0173] The core of the event triggering mechanism lies in achieving a comprehensive optimization of control performance and computing resource allocation through the design of reasonable triggering criteria. The triggering threshold conditions are defined in this step as shown in equations (29) and (30):

[0174] (29)

[0175] (30)

[0176] In equations (29) to (30), and This is the trigger threshold;

[0177] The expression for setting the event trigger threshold is:

[0178] (31)

[0179] (32)

[0180] In equations (31) and (32), E is a constant.

[0181] Online iteration ensures the system's adaptive control: loading the initial allowable control obtained from offline training. If the event triggering condition (15) or (16) is met, then training the evaluation network yields a value function that satisfies the policy evaluation condition, and training the execution network yields a control policy that satisfies the policy improvement condition; if in Internally satisfying the error condition Then update the optimal feed rate. The alumina concentration at the next moment is obtained by inputting it into the aluminum electrolysis process model. Otherwise, repeat the iterative training.

[0182] If the event triggering conditions (15) and (16) are not met, then calculate directly. The alumina concentration at the next moment is obtained by inputting it into the aluminum electrolysis process model. .

[0183] Step S7: Input the real-time data into the controller and output the optimal feeding amount for the next moment.

[0184] Based on the same idea, this invention also provides an event-triggered alumina concentration control system for aluminum electrolysis cells. The system includes an alumina concentration dynamic equilibrium model construction module, a model discretization module, an error module construction module, a controller construction module, an execution network training module, a trigger condition judgment module, a controller execution module, and a feed rate output module.

[0185] The alumina concentration dynamic equilibrium model construction module is used to analyze the alumina concentration dynamic equilibrium process in aluminum electrolysis and construct an alumina concentration dynamic equilibrium model.

[0186] The model discretization module is used to discretize the alumina concentration dynamic equilibrium model and construct an alumina dissolution discrete time model.

[0187] The error module construction module is used to convert the discrete-time model of alumina dissolution into an error model;

[0188] The controller construction module is used to construct an alumina concentration controller based on an error model; the controller includes an execution network, an error model, and an evaluation network.

[0189] The execution network training module is used to perform offline training on the execution network;

[0190] The trigger condition judgment module is used to collect the real-time data of the aluminum electrolysis cell at the current moment; it judges whether the real-time data of the material at the current moment meets the trigger condition based on the event triggering mechanism; if it does, the controller execution module is started; if it does not meet, the material feeding amount at the current moment is used as the material feeding amount at the next moment, and the material feeding amount output module is started directly.

[0191] The controller execution module is used to input real-time data into the controller to generate the optimal feeding amount for the next moment;

[0192] The material feeding output module is used to output the optimal material feeding amount for the next moment.

[0193] The system or device for executing the method in this embodiment of the invention can be a terminal or a server. The system includes a processor, a memory, and / or a transceiver, etc., and is connected via a communication bus. Each module can be implemented by a processor, a memory, and / or a transceiver, etc. The processor can be, but is not limited to, one or more microprocessors (MPUs), central processing units (CPUs), network processors (NPs), digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), and other programmable logic devices, discrete gates, transistor logic devices, discrete hardware components, etc., or can be configured to implement one or more integrated circuits of this invention. The processor can perform various functions by running or executing software programs in the memory and calling data in the memory. The memory includes Random Access Memory (RAM), Read-Only Memory (ROM), Electrically Erasable Programmable Read-Only Memory (EEPROM), Compact Disc Read-Only Memory (CD-ROM), and / or Non-Volatile Memory (NVM), etc. The transceiver is used to communicate with network devices or terminal devices, and includes a receiver and a transmitter. The memory and transceiver can be integrated with the processor or exist independently.

[0194] In the above embodiments, implementation can be achieved, in whole or in part, through software, hardware, firmware, or any combination thereof. When implemented in software, it can be implemented, in whole or in part, as a computer program product. The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, all or part of the processes or functions described in the embodiments of the present invention are generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via wired (e.g., coaxial cable, fiber optic, digital subscriber line (DSL)) or wireless (e.g., infrared, wireless, microwave, etc.) means.

[0195] It should also be noted that the event-triggered aluminum electrolytic cell alumina concentration control system described in this embodiment corresponds to the event-triggered aluminum electrolytic cell alumina concentration control method. The description and limitations of the method also apply to the system, and will not be repeated here.

[0196] As can be seen from the above technical solutions, the event-triggered alumina concentration control method and system for aluminum electrolysis cells provided in this invention, based on the characteristic analysis of the aluminum electrolysis alumina concentration control process, and considering the characteristics of "high complexity, multiple operating conditions, and strong coupling" in the aluminum electrolysis process control, adopts strategy iterative approximation of the Bellman equation to solve the problem of aluminum electrolysis alumina concentration tracking control in the finite time domain; for the initial allowable control required by the strategy iterative ADP, an offline learning scheme is designed to obtain the initial allowable control and the initial weights of the neural network required for iterative calculation, ensuring that the solved control sequence can meet the convergence of online tracking control; the introduction of an event-triggered mechanism significantly improves the utilization efficiency of the controller's computing resources, resulting in stronger adaptive capability, smaller overshoot, and higher control accuracy in the concentration control process.

[0197] It should be understood that, in various embodiments of the present invention, the order of the above-mentioned process numbers does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present invention.

[0198] The above description is merely a preferred embodiment of the present invention and an explanation of the technical principles employed, and is not intended to limit the scope of the claimed invention, but merely to illustrate preferred embodiments of the invention. Those skilled in the art should understand that the scope of the invention is not limited to the specific combinations of the above-described technical features, but should also cover other technical solutions formed by arbitrary combinations of the above-described technical features or their equivalents without departing from the inventive concept. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without inventive effort are within the scope of protection of the present invention.

Claims

1. A method for controlling alumina concentration in an aluminum electrolytic cell based on event triggering, characterized in that, The method includes the following steps: Step S1: Analyze the dynamic equilibrium process of alumina concentration in aluminum electrolysis and construct a dynamic equilibrium model of alumina concentration. Step S2: Discretize the dynamic equilibrium model of alumina concentration to construct a discrete time model of alumina dissolution. Step S3: Convert the discrete-time model of alumina dissolution into an error model; Step S4: Construct an alumina concentration controller based on an error model; the controller includes an execution network, an error model, and an evaluation network. Step S5: Train the execution network offline; Step S6: Collect real-time data of the aluminum electrolysis cell at the current moment; determine whether the real-time data of the material at the current moment meets the triggering conditions based on the event triggering mechanism; If satisfied, proceed to step S7; If the conditions are not met, output the current material feeding quantity as the optimal material feeding quantity for the next time step; Step S7: Input the real-time data into the controller and output the optimal feeding amount for the next moment.

2. The method according to claim 1, characterized in that, Step S1, which involves constructing a dynamic equilibrium model for alumina concentration, includes: Step S11: Analyze the dual kinetic characteristics of the alumina dissolution process in aluminum electrolysis and construct an alumina dissolution model; Step S12: Based on the alumina dissolution model and Faraday's law of electrolysis, a dynamic equilibrium model of alumina concentration is constructed.

3. The method according to claim 2, characterized in that, The formula for the alumina dissolution model is as follows: (1) (2) In equations (1) and (2), It is the mass fraction of alumina powder at time t. It is the solubility constant of alumina powder; It is the alumina feeding rate at time t. The proportion is the ratio of powder to powder; It is the mass fraction of alumina aggregates. It is the solubility constant of alumina aggregates; It is the electrolysis temperature; It is the reference electrolysis temperature; It is an empirical constant; This represents the mass of the electrolyte melt.

4. The method according to claim 3, characterized in that, The dynamic equilibrium model for alumina concentration is as follows: (3) In equation (3), Let be the alumina concentration at time t; Here is the molar mass of aluminum oxide; I represents current efficiency; I represents the series current. It is Faraday's constant; The number of electrons transferred.

5. The method according to claim 1, characterized in that, In step S2, when discretizing the dynamic equilibrium model of alumina concentration, alumina concentration is used as the main controlled variable, and the mass fraction of alumina powder and the mass fraction of alumina agglomerates are used as auxiliary controlled variables. The relationship between the main and auxiliary variables is quantified by controlling the weights of the three variables. The alumina feeding rate is used as the control variable.

6. The method according to claim 5, characterized in that, The discrete-time model for alumina dissolution is as follows: (4) In equation (4), for The state variables related to alumina concentration at time and The utility function state penalty quadratic form is ,in, and ; for Alumina feeding rate at all times; Electrolytic consumption during the alumina dissolution process and ; Represents the dynamic matrix of alumina dissolution. This represents the input matrix for alumina feeding.

7. The method according to claim 1, characterized in that, When converting step S3 into the error model, the reference trajectory for alumina concentration is defined as follows: The trajectory function that satisfies for: (5) The tracking error is The steady-state control variable at this point is defined as: (6) The error model is: (7) In equation (7), This represents the error control amount at time k; Define the utility function as follows: (8) In equation (8), satisfy and ; and This is a penalty factor for alumina concentration tracking error and feeding rate; Define the performance index function as follows: (9) The optimal performance index function is expressed as: (10) Optimal cost function The optimal error control quantity can be obtained by satisfying the HJB equation. for: (11)。 8. The method according to claim 1, characterized in that, The input to the execution network described in step S4 is Tracking error at any time The output is Optimal error control quantity at time 1 The input to the error model is Moment and The output is Tracking error at any time ; Evaluate the network input The output is Optimal performance index function at time 1 .

9. The method according to claim 1, characterized in that, In step S6, when determining whether the triggering condition is met, the event trigger time is defined as a monotonically increasing time sequence. At the triggering time Update the material feeding amount in the control system. The feeding control strategy is saved to the zero-order hold; feeding amount Represented as: (12) In equation (12), , ; At non-event-triggered times Zero-order hold control strategy until the next event is triggered; Define the error trigger for material feed update for: (13) In equation (13), express The tracking error of alumina concentration at any given time. This represents the tracking error at the s-th trigger moment; definition This is the error control input for the error system; the optimal feed rate is: (14) In equation (14), This represents the optimal feed rate at the s-th trigger time; The trigger threshold conditions are defined as shown in equations (15) and (16): (15) (16) In equations (15) and (16), and This is the trigger threshold; The expression for setting the event trigger threshold is: (17) (18) In equations (17) and (18), It is a constant; Online iteration ensures the system's adaptive control: loading the initial allowable control obtained from offline training. If the event triggering condition (15) or (16) is met, then training the evaluation network yields a value function that satisfies the policy evaluation condition, and training the execution network yields a control policy that satisfies the policy improvement condition; if in Internally satisfying the error condition Then update the optimal feed rate. The alumina concentration at the next moment is obtained by inputting it into the aluminum electrolysis process model. ; Otherwise, repeat the iterative training; If the event triggering conditions (15) and (16) are not met, then calculate directly. The alumina concentration at the next moment is obtained by inputting it into the aluminum electrolysis process model. .

10. An event-triggered alumina concentration control system for an aluminum electrolytic cell, characterized in that, The system includes an alumina concentration dynamic equilibrium model construction module, a model discretization module, an error module construction module, a controller construction module, an execution network training module, a trigger condition judgment module, a controller execution module, and a material feed output module; wherein... The alumina concentration dynamic equilibrium model construction module is used to analyze the alumina concentration dynamic equilibrium process in aluminum electrolysis and construct an alumina concentration dynamic equilibrium model. The model discretization module is used to discretize the alumina concentration dynamic equilibrium model and construct an alumina dissolution discrete time model. The error module construction module is used to convert the discrete-time model of alumina dissolution into an error model; The controller construction module is used to construct an alumina concentration controller based on an error model; the controller includes an execution network, an error model, and an evaluation network. The execution network training module is used to perform offline training on the execution network; The trigger condition judgment module is used to collect the real-time data of the aluminum electrolysis cell at the current moment; it judges whether the real-time data of the material at the current moment meets the trigger condition based on the event triggering mechanism; if it does, the controller execution module is started; if it does not meet, the material feeding amount at the current moment is used as the material feeding amount at the next moment, and the material feeding amount output module is started directly. The controller execution module is used to input real-time data into the controller to generate the optimal feeding amount for the next moment; The material feeding output module is used to output the optimal material feeding amount for the next moment.

Images