Efficient metal ion leaching system based on multi-parameter dynamic synergy and digital twinning

Abstract

The invention discloses an efficient metal ion leaching system based on multi-parameter dynamic collaboration and digital twinning, which is characterized in that a digital twinning system which is mapped with a physical leaching system in real time is constructed, a prospective instruction is generated by utilizing the predictive deduction capability of the digital twinning system, and the high-efficiency leaching of metal ions is realized through a self-adaptive adjustment algorithm. The instructions are converted into real-time adjustment of a dynamic weight distribution mechanism in a multi-parameter cooperative controller, so that predictive optimization and closed-loop dynamic adjustment of the leaching process are realized, and the unique architecture enables the whole control system to have the self-evolution capability of increasing the intelligence when the system is used; a robustness control module is arranged in the system, extreme working conditions such as raw material component fluctuation, sensor data abnormity and actuator faults can be dealt with, stable operation under the industrial environment is ensured, and benefited from the core innovation, under the optimal process parameters, the leaching efficiency of lithium and the leaching efficiency of iron stably exceed 96% and 94% respectively, and the leaching efficiency of lithium and iron is greatly improved. And meanwhile, the comprehensive energy consumption is reduced by 25-35%.

Description

Technical Field

[0001] This invention relates to the field of hydrometallurgical technology, specifically to a high-efficiency metal ion leaching system and method based on multi-parameter dynamic coordination and digital twins. It is particularly suitable for the synchronous, efficient, and intelligent recycling of waste lithium-ion batteries (including but not limited to LiFePO4, NCM, NCA, LCO) and mineral or metallurgical waste containing metals such as lithium, iron, cobalt, nickel, and manganese. Background Technology

[0002] Hydrometallurgy, as a traditional industrial field centered on physicochemical processes, has long relied on process improvements, equipment upgrades, and reagent optimization for technological development. This presents a significant gap between it and cutting-edge information technologies such as digital twins and reinforcement learning, which are based on data models and algorithms. Those skilled in the art typically lack the knowledge and motivation to deeply integrate these two distinct technological systems, and are even less likely to foresee the potentially disruptive effects of their combination.

[0003] In the prior art, some publicly available leaching methods focus only on optimizing a single process parameter, such as optimizing only the reaction temperature or the concentration of the leaching agent; other technical solutions, although considering multiple parameters, usually adopt simple linear combination or step-by-step control, failing to establish a dynamic synergistic optimization model between parameters, and lacking the ability to make forward-looking adjustments based on prediction. Their control logic is essentially still a passive response control based on real-time feedback.

[0004] In summary, existing leaching technologies have the following limitations:

[0005] 1) Parameter optimization is mostly static, single or linear combination, lacking systematic and synergistic consideration, and cannot cope with complex nonlinear reaction processes; and it is extremely sensitive to changes in raw material composition and lacks adaptability.

[0006] 2) Process control relies heavily on human experience, resulting in delayed response and large batch-to-batch efficiency fluctuations; when faced with extreme conditions such as sensor failure or actuator malfunction, there is a lack of effective fault handling and safety assurance mechanisms.

[0007] 3) High energy consumption, high processing costs, and a lack of overall energy efficiency optimization strategies;

[0008] 4) Lack of intelligent control systems with predictive and self-learning capabilities. Summary of the Invention

[0009] To solve the above-mentioned technical problems, the present invention adopts the following technical solution:

[0010] A high-efficiency metal ion leaching system based on multi-parameter dynamic coordination and digital twins is provided to construct a physical-virtual-algorithm closed-loop self-evolving architecture and achieve data-driven closed-loop optimization control. The system includes:

[0011] The online spectral monitoring module is used to collect process parameter information of the physical leaching reaction system in real time during the leaching process. The process parameter information includes metal ion concentration data, pH value and conductivity, so as to comprehensively reflect the changes in the chemical environment during the leaching process.

[0012] The leaching process digital twin module is communicatively connected to the online spectral monitoring module and is used for data information sent by the online spectral monitoring module. The high-fidelity neural network model performs simulation and deduction based on the acquired data information to predict and generate predictive control commands containing future parameter adjustment strategies.

[0013] The adaptive adjustment algorithm module based on deep reinforcement learning is communicatively connected to the leaching process digital twin module and the multi-parameter collaborative controller. It is used to receive the predictive control commands issued by the leaching process digital twin module and convert them into the current weight adjustment value.

[0014] The multi-parameter collaborative controller is communicatively connected to the adaptive adjustment algorithm module and the leaching process digital twin module. Its core is a collaborative optimization model with a dynamic weight allocation mechanism. It receives the current weight adjustment value and updates the weight of the collaborative optimization model in real time according to the current weight adjustment value (updating the weight in the collaborative optimization model of the core of the multi-parameter collaborative controller - i.e. the weight factor in the dynamic weight allocation mechanism) and outputs the final process parameter control command.

[0015] The intelligent energy management module is communicatively connected to the multi-parameter collaborative controller and the leaching process digital twin module. It is used to evaluate energy efficiency and feed the evaluation results back to the leaching process digital twin module for iteration of the high-fidelity neural network model.

[0016] In a preferred embodiment of the present invention, the dynamic weight allocation mechanism is a real-time weight update mechanism driven by predictive control commands of the leaching process digital twin module, and its mathematical model is expressed as follows: Where: η_total is the total leaching efficiency (or comprehensive optimization target), which is the weighted sum of the contribution values ​​of each process parameter, representing the overall optimization effect of the system;

[0017] w_T(t) is a dynamic weighting factor for the reaction temperature T. It is a function of time t and is not a fixed value. It is used to represent the weight of the temperature on the overall optimization at the current moment.

[0018] η_T is the contribution of reaction temperature T to the overall efficiency (or the corresponding efficiency component).

[0019] w_t(t) is the dynamic weighting factor of reaction time t. It is a function of time t and is used to represent the weight of the reaction time in the overall optimization at the current moment.

[0020] η_t is the contribution of reaction time t to the overall efficiency;

[0021] w_S / L(t) is the dynamic weighting factor of the solid-liquid ratio S / L. It is a function of time t and is used to represent the weight of the solid-liquid ratio in the overall optimization at the current moment.

[0022] η_S / L represents the contribution of the solid-liquid ratio (S / L) to the overall efficiency.

[0023] w_ω(t) is the dynamic weighting factor of the stirring speed (ω), which is a function of time t and is used to represent the weight of the stirring speed on the overall optimization at the current moment.

[0024] η_ω is the contribution of stirring speed (ω) to the overall efficiency;

[0025] In addition, the aforementioned weighting factor w_i(t) is not a fixed value. It is dynamically calculated by an adaptive adjustment algorithm based on the predictive instructions of the digital twin system. This mechanism enables the optimization focus to shift in real time based on future trends (such as reaction rate, energy consumption gradient, and ion concentration change trends).

[0026] In a preferred embodiment of the present invention, the "physical-virtual-algorithm closed-loop self-evolutionary architecture" is specifically manifested as follows: the high-fidelity neural network model is calibrated in real time based on the real-time acquired operating data; the accuracy of the digital twin model improves the training effect of the reinforcement learning algorithm; and the optimized control strategy makes the physical leaching reaction system run more stably and generate higher quality data. The three form a positive feedback closed loop, which enables the overall system performance to continuously improve without human intervention. The operating data includes comprehensive process data such as metal ion concentration, temperature, time, solid-liquid ratio, stirring speed, pH value, conductivity, and power.

[0027] In a preferred embodiment of the present invention, the multi-parameter collaborative controller, the digital twin system, and the intelligent energy management system constitute a closed-loop, self-evolving optimization architecture: the digital twin system generates predictive optimization strategies, the multi-parameter collaborative controller executes short-term precise control, and the intelligent energy management system evaluates long-term energy efficiency, with the evaluation results fed back to the digital twin system for model correction and iteration; the closed-loop, self-evolving optimization architecture enables the system to autonomously optimize its internal digital twin model and control strategies through continuous "physical-virtual-physical" data interaction without human intervention, thereby achieving continuous improvement in processing efficiency and energy efficiency.

[0028] In a preferred embodiment of the present invention, a robust control module is further included. The robust control module acquires real-time sensor data and identifies data anomalies by comparing the confidence intervals of the real-time sensor data with the predicted values ​​in the digital twin module of the leaching process. When the real-time sensor data is not within the confidence interval, it is judged as an anomaly, the abnormal data point is temporarily ignored, and short-term maintenance control is performed based on the previous valid data and the digital twin prediction, while triggering an alarm.

[0029] The real-time sensor data includes other key process parameters such as metal ion concentration, temperature, pressure, and pH value. The predicted value refers to the estimated value of the system state (such as metal ion concentration and temperature) at the current or future moment, calculated by the high-fidelity neural network model of the leaching process digital twin module based on real-time data.

[0030] There is a progressive foundation-decision-execution relationship between predicted values, future parameter adjustment strategies, and predictive control commands. Predicted values ​​are the basis for generating strategies, strategies are the content of commands, and commands are the final form sent to the controller.

[0031] Specifically:

[0032] Predicted values ​​(basis / foundation): The digital twin module calculates "predicted values" (such as predicting future ion concentrations) based on real-time data.

[0033] Future parameter adjustment strategy (decision): Based on these "predictions", the digital twin module analyzes trends and formulates a plan to optimize the reaction process, namely the "future parameter adjustment strategy" (e.g., if the reaction rate is predicted to decrease, the strategy is to "extend the reaction time" or "increase the temperature").

[0034] Predictive control instructions (execution signals): The above "strategy" is converted into specific control signals, namely "predictive control instructions", and sent to the multi-parameter cooperative controller for execution (e.g., sending specific instructions containing "increase the weight of w_t").

[0035] In a preferred embodiment of the present invention, the robustness control module can automatically trigger a safety shutdown procedure when it detects that the process parameters exceed the safe operating window. The safety shutdown procedure includes cutting off the heating source, stopping stirring, and recording a fault log. The process parameters include temperature and pressure. The process parameters are not entirely the same as sensor data. Although from a physical hardware perspective, temperature and pressure are also data collected by sensors, for the sake of strict logical distinction: temperature and pressure are separately categorized as process parameters for triggering shutdown; while sensor data is mainly categorized as "metal ion concentration, etc." for data quality verification. In this patent, the two are data sets for different purposes.

[0036] In a preferred embodiment of the present invention, the response time of the adaptive adjustment algorithm module to dynamically adjust the weight value is less than 1 second.

[0037] In a preferred embodiment of the present invention, the method for constructing the digital twin module of the leaching process includes: training a high-fidelity neural network model using historical experimental data and physicochemical mechanisms; receiving data from an online spectral monitoring module and a multi-parameter collaborative controller in real time to maintain synchronization with the state of the physical leaching reaction system; rapidly simulating the impact of different parameter combinations on the final leaching efficiency and total energy consumption in a virtual environment to predict and generate the optimal future parameter adjustment strategy; wherein the prediction time window is 0.5-2 hours in the future; and the accuracy standard for prediction and extrapolation is: the average absolute error between the predicted value and the actual measured value of the metal ion concentration in the next hour is less than 5%.

[0038] In a preferred embodiment of the present invention, the intelligent energy management module adopts a combined optimization strategy of genetic algorithm and particle swarm optimization algorithm, and uses the long-term energy consumption prediction provided by the digital twin system as the objective function for optimization.

[0039] A leaching control method for a high-efficiency metal ion leaching system based on multi-parameter dynamic coordination and digital twins, comprising the following steps:

[0040] Step S1: The online spectral monitoring module collects metal ion concentration data in real time during the leaching process;

[0041] Step S2: The digital twin module of the leaching process performs simulation and deduction based on the acquired metal ion concentration data to predict and generate future parameter adjustment strategies, and sends predictive control commands with future parameter adjustment strategies.

[0042] Step S3: The adaptive adjustment algorithm module calculates the current weight adjustment value of the collaborative optimization model based on the future parameter adjustment strategy in the predictive control command;

[0043] Step S4: The multi-parameter collaborative controller receives the current weight adjustment value, updates the dynamic weight allocation mechanism of its internal collaborative optimization model in real time, calculates the final process parameters, and outputs the control command with the final process parameters to the physical leaching reaction system.

[0044] Step S5: The intelligent energy management system evaluates energy efficiency in real time and feeds the evaluation results back to the leaching process digital twin module for the next round of model iteration and control.

[0045] The beneficial effects of this invention are as follows: By constructing a digital twin system that is mapped in real time to the physical leaching system, the system utilizes its predictive and extrapolation capabilities to generate forward-looking instructions. These instructions are then transformed into real-time adjustments to the "dynamic weight allocation mechanism" in the multi-parameter collaborative controller via an adaptive adjustment algorithm. This achieves predictive optimization and closed-loop dynamic adjustment of the leaching process. This unique architecture enables the entire control system to have a self-evolving capability that "gets smarter with use." The system can cope with extreme conditions such as fluctuations in raw material composition, abnormal sensor data, and actuator failures, ensuring stable operation in industrial environments. It completely solves the problems of rigid parameter optimization, lagging process control, high energy consumption, and inability to adapt to raw material fluctuations in traditional leaching technologies. Attached Figure Description

[0046] 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 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, wherein:

[0047] Figure 1 This is a schematic diagram of a preferred embodiment of the high-efficiency metal ion leaching system based on multi-parameter dynamic coordination and digital twin of the present invention. Detailed Implementation

[0048] The technical solutions in the embodiments of the present invention will be clearly and completely described below. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of them. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0049] See Figure 1 The embodiments of the present invention include:

[0050] A high-efficiency metal ion leaching system based on multi-parameter dynamic coordination and digital twins is proposed. Its core lies in the construction of a data-driven closed-loop optimization control loop, in which each module plays a clear and coordinated role.

[0051] 1. Online Spectral Monitoring Module: Serving as the system's "sensory organ," it employs an ICP-OES online detection system with a detection frequency of 1 time / minute, detection limits of Li 0.01 ppm and Fe 0.05 ppm, and an RSD of <1%. Its core function is to acquire real-time data on metal ion concentrations during the leaching process of the physical reaction system (reactor) and use this data as the primary input to drive the entire optimized loop.

[0052] 2. Digital Twin Module for Leaching Process: As the "brain" of the system, it includes a high-fidelity neural network model that is mapped 1:1 to the physical leaching reactor. The core function of this model is to receive real-time detection (metal ion concentration) data from the online spectral monitoring module and perform simulations based on this data to generate predictive instructions containing future parameter adjustment strategies, providing a decision-making basis for the "dynamic weight allocation mechanism".

[0053] The construction of the digital twin module for the leaching process is based on: a) historical experimental data; b) the physicochemical mechanism of the leaching reaction.

[0054] In a preferred embodiment, the historical experimental data includes at least 1000 sets of valid data points, each set of data points consisting of parameters in the following dimensions:

[0055] Input parameters include reaction temperature (range: 120-160℃), reaction time (range: 4-12h), solid-liquid ratio (range: 1:50-1:150g / g), stirring speed (range: 200-600r / min), type and amount of DES, particle size distribution of raw materials, and initial content of Li and Fe in raw materials;

[0056] The output parameters include Li ion concentration, Fe ion concentration, solution pH, conductivity, real-time power, and cumulative energy consumption at different time points.

[0057] In a preferred embodiment, the high-fidelity neural network model employs a Long Short-Term Memory (LSTM) network model to capture the temporal dynamics of the leaching process. Specifically, the network structure is as follows: the input layer has 8 nodes, corresponding to the input temperature, time, solid-liquid ratio, stirring speed, DES dosage, and the previous time-series Li concentration, Fe concentration, and pH value; the network contains 3 hidden layers, each with 128 neurons and an activation function of tanh; the output layer has 4 nodes, corresponding to the predicted Li concentration, Fe concentration, power, and pH value for the next hour.

[0058] In a preferred embodiment, the training parameters of the LSTM model are set as follows: 5000 iterations, batch size of 32, Adam optimizer, initial learning rate of 0.001, learning rate decay strategy (decaying to 80% of the original learning rate every 1000 iterations), and training termination condition is that the validation set loss function (L_total) does not decrease for 50 consecutive iterations.

[0059] Regarding the integration of physicochemical mechanisms and models, this invention employs physical information neural network technology for fusion. Specifically, when training the LSTM model, its loss function consists of two parts: .

[0060] Where L_data is the mean squared error loss, which measures the difference between the model's predicted values ​​and the historical experimental data values, and λ_data is the weighting coefficient (or balancing coefficient) of the mean squared error loss (Ldata). In the total loss function, it is used to adjust the proportion of "data fitting error" in the total loss, and determines the degree of importance the model attaches to fitting historical experimental data during the training process.

[0061] L_physics represents the physical constraint loss, and its design is based on the kinetic equations of the contraction kernel model of the leaching reaction, i.e. , where k(T) is the rate constant related to temperature T, f(C_Li) is a function term related to concentration, and dC_Li / dt is the first derivative of lithium ion concentration (CLi) with respect to time (t). In the kinetic context of the leaching reaction, it represents the rate of change of lithium ion concentration, i.e., the leaching reaction rate of lithium ions.

[0062] λ_physics is the weighting coefficient (or balance coefficient) of the physical constraint loss (Lphysics). In the total loss function, it is used to adjust the proportion of "physical constraint error" in the total loss, and determines the strength of the model's adherence to and constraint of the physicochemical mechanism of leaching reaction (such as kinetic equations) during training.

[0063] By minimizing L_physics, the neural network's predictions are forced to follow known physical and chemical laws, thereby greatly improving the model's generalization ability and prediction accuracy under unseen conditions.

[0064] 3. Adaptive adjustment algorithm module based on deep reinforcement learning: As the "nerve center" of the system, its core function is to connect the "brain" and the "cerebellum", receive predictive control instructions from the digital twin module of the leaching process, and convert them into specific weight adjustment values ​​that the multi-parameter collaborative controller can understand, thereby driving the operation of the "dynamic weight allocation mechanism".

[0065] In a preferred embodiment, the reward function R is designed as follows: .

[0066] Where Δη_total is the increment of leaching efficiency, and η_target refers to the target leaching efficiency in the reward function formula. In this context, it serves as the denominator of the baseline value, used to normalize the increment of leaching efficiency (Δηtotal) to assess the extent to which the current efficiency improvement has met the target.

[0067] ΔE_total represents the increment of energy consumption, and E_target refers to the target energy consumption, as shown in the reward function formula. In this context, it serves as the denominator of the baseline value, used to normalize the increment of energy consumption (ΔEtotal) to assess the extent of the current energy consumption change relative to the target.

[0068] Penalty is a penalty for exceeding safe operating limits (such as exceeding temperature or pressure limits).

[0069] α, β, and γ are weighting coefficients (e.g., α = 0.6, β = 0.3, γ = 0.1). Its state space is defined as follows: Action space is defined as .

[0070] In the State Space (S), the parameters are defined as follows: T represents the reaction temperature, t represents the reaction time, S / L represents the solid-liquid ratio, ω represents the stirring speed, C_Li represents the lithium ion concentration, C_Fe represents the iron ion concentration, dC_Li / dt represents the rate of change of lithium ion concentration (i.e., the reaction rate of lithium ions), dC_Fe / dt represents the rate of change of iron ion concentration (i.e., the reaction rate of iron ions), and Power represents the real-time power.

[0071] In the Action Space (A), the parameters are defined as follows: ΔT represents the temperature adjustment value (i.e., the increment for adjusting the reaction temperature), Δt represents the time adjustment value (i.e., the increment for adjusting the reaction time), and Δω represents the stirring speed adjustment value (i.e., the increment for adjusting the stirring speed).

[0072] Etotal refers to total energy consumption, while ΔEtotal refers to the increment of energy consumption, i.e., the change in energy consumption during the control process. ΔEtotal is the change (or difference) in Etotal. In the reward function calculation of reinforcement learning, ΔEtotal is used to measure the increase or decrease in the total energy consumption (Etotal) of the system when transitioning from the previous state to the current state.

[0073] In a preferred embodiment, the adaptive adjustment algorithm module performs weight adjustment in less than 1 second, ensuring real-time control.

[0074] In a preferred embodiment, the training process of the reinforcement learning algorithm is as follows: the training environment is built based on a digital twin system, the training steps are 1 million steps, and the exploration rate is... The reward value decreases linearly from 0.9 to 0.1, the experience replay buffer has a capacity of 100,000 records, the target network is updated every 1,000 steps, and the training is considered complete when the average reward value of 100 consecutive episodes is ≥0.8 (the maximum value of the reward function is 1).

[0075] 4. Multi-parameter Cooperative Controller: As the "cerebellum" of the system, it communicates with the adaptive adjustment algorithm module and is responsible for precise execution. The multi-parameter cooperative controller is the core control unit of this invention. Its innovation lies in the fact that the multi-parameter cooperative controller adopts a cooperative optimization model, and the weights of the cooperative optimization model are not fixed, but are updated in real time according to the weight adjustment values ​​received from the adaptive adjustment algorithm module, realizing a "dynamic weight allocation mechanism" and outputting the final process parameter control instructions.

[0076] The input variables of the multi-parameter collaborative controller are: temperature, time, solid-liquid ratio, stirring speed, pH value, and conductivity.

[0077] Output variables of the multi-parameter collaborative controller: lithium concentration, iron concentration, energy consumption, and efficiency prediction.

[0078] The control accuracy of the multi-parameter collaborative controller is: temperature ±0.5℃, time ±0.1h, and solid-liquid ratio ±2g / g.

[0079] The "dynamic weight allocation mechanism" is a real-time weight update mechanism driven by predictive instructions from the digital twin module of the leaching process. Its mathematical model is expressed as follows: .

[0080] η_total is the total leaching efficiency (or comprehensive optimization target), which is the weighted sum of the contribution values ​​of each process parameter and represents the overall optimization effect of the system.

[0081] w_T(t) is a dynamic weighting factor for the reaction temperature T. It is a function of time t and is not a fixed value. It is used to represent the weight of the temperature on the overall optimization at the current moment.

[0082] η_T is the contribution of reaction temperature T to the overall efficiency (or the corresponding efficiency component).

[0083] w_t(t) is the dynamic weighting factor of reaction time t. It is a function of time t and is used to represent the weight of the reaction time in the overall optimization at the current moment.

[0084] η_t is the contribution of reaction time t to the overall efficiency;

[0085] w_S / L(t) is the dynamic weighting factor of the solid-liquid ratio S / L. It is a function of time t and is used to represent the weight of the solid-liquid ratio in the overall optimization at the current moment.

[0086] η_S / L represents the contribution of the solid-liquid ratio (S / L) to the overall efficiency.

[0087] w_ω(t) is the dynamic weighting factor of the stirring speed (ω), which is a function of time t and is used to represent the weight of the stirring speed on the overall optimization at the current moment.

[0088] η_ω is the contribution of stirring speed (ω) to the overall efficiency;

[0089] In addition, the aforementioned weighting factor w_i(t) is not a fixed value. It is dynamically calculated by an adaptive adjustment algorithm based on the predictive instructions of the digital twin system. This mechanism enables the optimization focus to shift in real time based on future trends (such as reaction rate, energy consumption gradient, and ion concentration change trends).

[0090] The weighting factor w_i(t) is not a fixed value, but is dynamically calculated by the adaptive adjustment algorithm based on key indicators such as real-time reaction rate, energy consumption gradient, and ion concentration change trend, so as to realize the real-time shift of the optimization focus.

[0091] In a preferred embodiment, the optimal process parameters are: temperature 140±2℃, time 8±0.5 hours, solid-liquid ratio 1:100±5g / g, stirring speed 400±50r / min, pH value 3.0-3.5, and reaction pressure at atmospheric pressure.

[0092] In a preferred embodiment, the collaborative optimization model employs a fuzzy neural network (FNN) algorithm, whose network structure includes 5 input layer nodes, 3 hidden layers (each containing 20-30 neurons, with ReLU or Sigmoid activation functions) and 2 output layer nodes.

[0093] 5. Intelligent Energy Management Module: Communicating with the multi-parameter collaborative controller and the leaching process digital twin module, the intelligent energy management module acts as the system's "energy monitor." Its core function is to evaluate energy efficiency and feed the evaluation results back to the leaching process digital twin module for high-fidelity neural network model iteration. It is a key component in forming the "physical-virtual-algorithm closed-loop self-evolving architecture."

[0094] In a preferred embodiment, the intelligent energy management module has the following features: power monitoring: 0-10kW, accuracy ±0.01kW; energy consumption calculation: real-time accumulation, accuracy ±0.1kWh; optimization algorithm: genetic algorithm (GA) + particle swarm optimization (PSO).

[0095] In a preferred embodiment, the parameters of the genetic algorithm are set as follows: population size 50-100, crossover probability 0.6-0.8, mutation probability 0.01-0.05; the parameters of the particle swarm optimization algorithm are set as follows: number of particles 50-100, inertia weight 0.4-0.9, learning factor c1=c2=2.0.

[0096] 6. System Data Flow and Hardware Integration Architecture: To ensure the manufacturability and high reliability of the system of this invention, this invention defines a complete data flow and hardware integration architecture to ensure the real-time, accuracy and standardization of information interaction between modules.

[0097] Data transmission protocol and format: The online spectral monitoring module (ICP-OES) serves as the core data source. Its real-time data stream is published to the multi-parameter collaborative controller via industrial Ethernet using the OPCUA (Unified Architecture) protocol. The data format adopts the lightweight JSON (JavaScript Object Notation) format. A typical data packet example is as follows: {"timestamp":"2023-10-27T10:00:00Z","Li_conc_ppm":1.23,"Fe_conc_ppm":0.98,"pH":3.1,"Conductivity_uS_cm":1500}. The end-to-end transmission delay from ICP-OES data acquisition to controller reception is strictly controlled within 100 milliseconds to meet real-time control requirements.

[0098] Hardware Interface Specifications: The multi-parameter co-controller connects to the actuators of the physical leaching reaction system using an industry-standard interface. For the heater, the multi-parameter co-controller sends commands to the SCR power regulator via Modbus TCP / IP protocol to precisely control the heating power. For the stirring motor, the multi-parameter co-controller outputs a 4-20mA current signal to the VFD via an analog output (AO) module, thereby linearly adjusting the stirring speed. All hardware devices are connected via an industrial switch, forming a unified control network.

[0099] Internal data bus: The multi-parameter collaborative controller, the leaching process digital twin module, and the intelligent energy management module are deployed on the same high-performance industrial computer (IPC) or server. The three exchange data through a high-speed internal bus (such as PCIe) or gigabit Ethernet to ensure that the simulation results and optimization strategies of the leaching process digital twin module can be transmitted to the multi-parameter collaborative controller with a microsecond delay.

[0100] This application does not simply superimpose algorithms such as digital twins and reinforcement learning onto conventional hardware. Instead, through innovative system architecture design, it constructs two core mechanisms: a "dynamic weight allocation mechanism" and a "physical-virtual-algorithm positive feedback loop." This addresses a long-standing technical bias in the field that was considered difficult to overcome through traditional methods, and generates a synergistic effect of "1+1>2." The core mechanisms include:

[0101] (A) Dynamic weight allocation mechanism

[0102] This invention pioneers a dynamic weight allocation mechanism driven by "predictive instructions." Unlike traditional control methods where weights are fixed or adjusted only based on the current state with lag, the weight adjustment in this invention is directly derived from the prediction of future states by the digital twin module of the leaching process. For example, instead of increasing the time weight only after detecting a decrease in the reaction rate, the system fine-tunes the weight in advance when the digital twin predicts that "the reaction rate will begin to decrease in 30 minutes," achieving true "predictive" control. This mechanism allows the optimization focus to be shifted to the most critical parameters in real time and with precision, resulting in control accuracy and efficiency far exceeding any existing technology.

[0103] (B) Physical-Virtual-Algorithm Closed-Loop Self-Evolving Architecture

[0104] This invention constructs a unique "physical-virtual-algorithm" positive feedback loop. Based on the operational data of the physical leaching reaction system, a high-fidelity neural network model can be continuously calibrated, making it increasingly accurate. The accuracy of the digital twin model, in turn, improves the training effect of the reinforcement learning algorithm. The optimized control strategy makes the physical system more stable and the data quality higher. This forms a positive feedback loop among the three, giving the entire system a self-evolving capability that "gets smarter with use." This is a completely new system form, and its long-term benefits and adaptability are unmatched by any static combination of technologies.

[0105] This application, by organically integrating the two core innovations mentioned above, can produce the following beneficial effects:

[0106] 1. A paradigm shift from “passive response” to “proactive prediction”.

[0107] The digital twin module for the leaching process in this invention provides a zero-risk, highly accelerated virtual training ground for reinforcement learning algorithms. Reinforcement learning algorithms can undergo millions of hypothesis-and-deduction trials in the digital twin environment, rapidly learning optimal control strategies, and then outputting these strategies (predictive instructions) to the physical leaching reaction system. This combination enables the system to no longer passively react to the current state, but rather to intervene in advance based on predictions of future trends, achieving a fundamental leap in control strategies.

[0108] 2. Refined control from "static optimization" to "dynamic focus".

[0109] This invention employs a dynamic weight allocation mechanism. The adjustment of dynamic weights is not based on current monitoring data, but directly on the digital twin system's prediction of future states. This "predictive instruction-driven weight adjustment" mechanism allows the optimization focus to be shifted to the most critical parameters in real time and with precision, achieving control accuracy and efficiency far exceeding any single technology or simple combination.

[0110] 3. The life cycle leap from "fixed system" to "self-evolving intelligent agent".

[0111] This invention employs a closed-loop self-evolving architecture of physics-virtualization-algorithm. It continuously calibrates a high-fidelity neural network model based on operational data from the physical leaching reaction system, making it increasingly accurate. The accuracy of the digital twin model, in turn, enhances the training effect of the reinforcement learning algorithm. The optimized control strategy makes the physical system more stable and the data quality higher. This forms a positive feedback loop between physics, virtualization, and algorithm, enabling the entire system to possess a self-evolving capability that becomes "smarter with use."

[0112] 4. Wide range of applications.

[0113] The core architecture of this invention is highly versatile. Although the embodiment uses LiFePO4 as an example, its principle is equally applicable to the leaching process of other chemical systems. For example, when processing ternary lithium batteries (NCM), it is only necessary to add the leaching kinetic data of Ni, Co, and Mn to the training data of the digital twin system and adjust the reward function of reinforcement learning to simultaneously optimize the leaching efficiency of these three metals. Therefore, this invention can be widely applied to the leaching of waste lithium iron phosphate batteries, ternary lithium batteries (NCM523, 622, 811, etc.), lithium cobalt oxide batteries, and secondary resources such as metal ores and smelting slags containing lithium, iron, cobalt, nickel, and manganese.

[0114] Regarding raw material specifications, in a preferred embodiment, for waste LiFePO4 powder, the system requires that its active material content be no less than 85% (by weight). Major impurities such as copper should have a content of less than 0.5%, and aluminum should have a content of less than 1.0%. When the raw material composition fluctuates within this range, the digital twin system can effectively predict and optimize through its generalization capabilities. When the impurity content exceeds this range, the system will issue an early warning and may adjust to more conservative process parameters.

[0115] 5. Robustness under extreme operating conditions.

[0116] To ensure stable operation in complex industrial environments, this invention designs a dedicated robust control mechanism.

[0117] a. Addressing Fluctuations in Raw Material Composition: When online monitoring or raw material scanning detects significant changes in raw material composition (such as impurity content), the digital twin module of the leaching process immediately reassesses the feasibility of the current process path and deduces new optimization strategies. For example, if excessive copper content is detected, the system may predict that copper leaching will consume more acid and strategically fine-tune the pH or time weight of subsequent reactions accordingly to avoid a decrease in the leaching efficiency of the main metal.

[0118] b. Handling Sensor Data Anomalies: The robust control module continuously monitors the data quality of the online spectral monitoring module and compares the real-time data with the confidence interval of the predicted values ​​in the leaching process digital twin module (hereinafter referred to as the prediction confidence interval). If a data point deviates from the prediction confidence interval (e.g., more than 3 standard deviations), the system will mark it as an outlier, temporarily ignore the data point, and perform short-term maintenance control based on the valid data from the previous period and the prediction trend of the digital twin, while simultaneously issuing a sensor calibration or maintenance alarm to the operator.

[0119] c. Handling actuator malfunctions or process parameter overruns: The system has a strict safety operation window. Once a critical process parameter (such as reactor temperature or pressure) is detected to exceed a preset safety threshold (e.g., temperature exceeding 150°C), the robust control module will immediately trigger a safety shutdown procedure to ensure the safety and effectiveness of production. The safety shutdown procedure includes, but is not limited to: immediately cutting off all heater power, reducing the stirring speed to a safe value, opening the emergency cooling circuit, and displaying a clear fault type and location on the human-machine interface (HMI), while simultaneously recording a detailed fault log for post-event analysis.

[0120] Example 1

[0121] This embodiment demonstrates the processing of 100g of pretreated LiFePO4 powder using this system. Initial parameter settings: temperature 140℃, time 8 hours, solid-liquid ratio 1:100g / g, stirring speed 400r / min, DES dosage 10L.

[0122] System operation process:

[0123] 1. Upon system startup, hardware self-test and network link establishment are performed first. ICP-OES is successfully registered to the multi-parameter collaborative controller via the OPCUA protocol. The leaching process digital twin module (using a 3-layer LSTM network with 128 neurons per layer, 8 input nodes, and 4 output nodes) is initialized synchronously with the physical leaching reaction system.

[0124] 2. In the first two hours, the lithium concentration rose rapidly. The ICP-OES sent concentration data to the multi-parameter collaborative controller every minute via JSON data packets, with an average end-to-end latency of 75 milliseconds. This data was input into the leaching process digital twin module in real time by the controller. When the leaching process digital twin module predicted that the current parameter combination was the optimal path, it generated predictive control commands based on the current parameter combination and sent them to the adaptive adjustment algorithm module, which converted them into weight adjustment values ​​to maintain the status quo. The multi-parameter collaborative controller (using an FNN with the ReLU activation function) received the weight adjustment values ​​and maintained the original parameters. w_T in the collaborative optimization model remained at a high value (0.9), reflecting the operation of the "dynamic weight allocation mechanism".

[0125] 3. In the third hour, online monitoring showed a decrease in the rate of increase in lithium concentration. The leaching process digital twin module immediately extrapolated based on the current detection parameters and predicted that if the current situation was maintained, the efficiency plateau (94%) would be reached in 8.5 hours. The leaching process digital twin module simulated a scheme (strategy) to extend the reaction by 0.3 hours, predicting that the efficiency of this scheme could reach 95.5%. Therefore, it determined that the parameter combination of this scheme was the optimal path and generated corresponding predictive control instructions based on the optimal path parameter combination. These predictive control instructions were sent to the controller, and the adaptive adjustment algorithm completed the weight calculation within 0.8 seconds, increasing the weight of w_t from 0.1 to 0.25. At the same time, the controller automatically extended the reaction timer by 0.3 hours. The above process clearly demonstrates the process of "predictive control instructions driving weight adjustment".

[0126] 4. In the 5th hour, an abnormal increase in energy consumption was detected. The digital twin module of the leaching process diagnosed the anomaly as localized overheating and derived a strategy to reduce the temperature to 138°C. This strategy predicted a 5% reduction in energy consumption with a less than 0.2% impact on the final efficiency. Based on this strategy, a predictive control command was generated. The controller received the predictive control command and adopted the strategy, sending a new power setpoint to the heater's SCR via the Modbus TCP / IP protocol, allowing the temperature to steadily decrease to 138°C.

[0127] 5. At the 7th hour, the iron concentration reached equilibrium. The digital twin module of the leaching process predicted that subsequent reactions would have minimal impact on iron leaching efficiency but would continuously increase energy consumption. Therefore, a scheme to end the reaction 0.2 hours earlier was generated, and a predictive control command was generated based on this scheme. The controller issued a stop command based on the predictive control command, reducing the speed of the stirring motor to 0 via a 4-20mA signal and cutting off the power to the heater.

[0128] Final results: Lithium leaching efficiency 96.2%, iron leaching efficiency 94.5%, total energy consumption 2.65 kWh, processing time 8.1 hours. The predictive MAE of the digital twin system in this embodiment was 3.2% for Li concentration and 4.1% for Fe concentration.

[0129] Example 2

[0130] When processing different batches of raw materials (200g, 500g, 1000g), the system rapidly scans the raw materials and inputs the initial data into the leaching process digital twin module. The leaching process digital twin module then generates differentiated initial optimization paths for each batch.

[0131] 200g: Digital twin prediction of high mass transfer efficiency, strategy is "high temperature and short time", set temperature 141℃, time 7.8 hours, stirring speed 420r / min.

[0132] 500g: Digital twin prediction shows a mass transfer bottleneck. The strategy is to "enhance stirring". The temperature is set at 139℃, the time is 8.2 hours, and the stirring speed is 380r / min (to increase turbulence).

[0133] 1000g: Digital twin predicts high thermal inertia, so the strategy is "stable temperature control", setting the temperature to 140℃, the time to 8.5 hours, and the stirring speed to 400r / min.

[0134] The three batches of treatment showed extremely high consistency: lithium leaching efficiency was 95.8-96.5%, and iron leaching efficiency was 94.1-94.8%, demonstrating the system's strong adaptability.

[0135] Example 3

[0136] System robustness verification under extreme conditions

[0137] Scenario A: Raw material impurities exceed limits. A batch of LiFePO4 raw material with a copper content of 1.2% was processed (exceeding the conventional upper limit of 0.5%). After system startup, the raw material scanning module identified the high copper content and issued an early warning to the leaching process digital twin module. The leaching process digital twin module immediately performed a simulation, predicting that copper leaching would additionally consume the effective components in the DES, leading to a decrease in the Li leaching rate in the later stages. Based on this, the system automatically adjusted its initial strategy: increasing the DES dosage by 5% and setting a higher temperature weight w_T in the early stages of the reaction to accelerate the reaction process. Ultimately, despite the excessive raw material impurities, the lithium leaching efficiency still reached 95.1%, demonstrating the system's adaptability to raw material fluctuations.

[0138] Scenario B: Sensor data anomaly. At the 4th hour of the reaction, the ICP-OES system experienced a brief blockage in the injection tubing, sending an abnormal spike signal of 50 ppm Li concentration (previously 1.5 ppm). The robustness control module detected that this value far exceeded the 1.6 ± 0.2 ppm confidence interval predicted by the digital twin system and immediately marked it as invalid data. The system did not malfunction as a result, but continued to operate stably based on the previous valid data and predicted trend. Simultaneously, the HMI displayed an alarm: "ICP-OES data anomaly, please check the injection system." After 30 seconds, the sensor returned to normal, and the system reconnected to the real-time data stream.

[0139] Scenario C: Actuator malfunction leads to temperature exceeding limits. Due to a failure in the heater power regulator (SCR), the reactor temperature continued to rise to 146℃, exceeding the safety window of 140±2℃. Upon detecting the temperature exceeding the limit, the robust control module triggered a safety shutdown procedure within 0.1 seconds: immediately cutting off the SCR main power supply, stopping heating, simultaneously reducing the stirring speed to 100 r / min to prevent localized overheating, and initiating cooling water circulation. The system shut down safely, effectively preventing potential safety accidents and equipment damage.

[0140] Example 4: Verification of generalization ability for different raw material types

[0141] To verify the broad applicability of the system of this invention, we applied it to different types of waste lithium battery cathode materials.

[0142] Scenario A: Processing NCM811 ternary lithium batteries. This system was used to process 200g of pretreated NCM811 powder. The initial parameter window was adjusted to a temperature of 150±2℃ and a time of 9±0.5 hours. After system operation, the digital twin system successfully predicted and optimized the simultaneous leaching process of four metals: Li, Co, Ni, and Mn. The final results were: lithium leaching efficiency 95.7%, cobalt leaching efficiency 94.3%, nickel leaching efficiency 93.8%, and manganese leaching efficiency 92.5%, with an overall energy consumption reduction of 28%. The average MAE (Modular Energy Estimation) for the predicted concentrations of the four metal ions by the digital twin system in this embodiment was measured to be 4.3%.

[0143] Scenario B: Processing lithium cobalt oxide batteries. This system was used to process 150g of pretreated LCO powder, with initial parameter windows adjusted to a temperature of 145±2℃ and a time of 7±0.5 hours. The system also exhibited excellent performance, with final results showing a lithium leaching efficiency of 96.5%, a cobalt leaching efficiency of 95.2%, and a 30% reduction in overall energy consumption. The digital twin system's predicted MAE for Li and Co concentrations were 3.8% and 4.0%, respectively. These results fully demonstrate the powerful generalization ability and adaptability of the system to different chemical systems.

[0144] The beneficial effects of this invention are:

[0145] 1. It has achieved a paradigm shift from "post-event control" to "pre-event prediction" and has the ability to self-evolve.

[0146] Through deep synergy between digital twins and reinforcement learning, the "physical-virtual-algorithm closed-loop self-evolving architecture" constructed in this invention enables the system not only to anticipate the future, avoid risks, and proactively seek optimization, but also to continuously learn and evolve during operation, becoming smarter with use. This forward-looking capability allows the system to explore the entire parameter domain in virtual space, and combined with a "dynamic weight allocation mechanism," guides the physical system to always evolve along the path leading to the global optimum, thereby producing a synergistic effect of "1+1>2".

[0147] 2. Achieved excellent and stable performance indicators.

[0148] Thanks to the refined control of the "dynamic weight allocation mechanism" and the continuous optimization of the "closed-loop self-evolution architecture", its direct results are reflected in its excellent and stable performance: under optimal conditions, the leaching efficiency of lithium and iron can be stably reached above 96% and 94% respectively, and the consistency between different batches is extremely high, which completely solves the pain point of large efficiency fluctuations in traditional technology.

[0149] 3. It combines ultimate greenness and economy.

[0150] Under mild reaction conditions, the system achieves ultimate energy efficiency through the collaboration of intelligent energy management and digital twins, reducing overall energy consumption by 25-35%, significantly reducing processing costs, and is environmentally friendly.

[0151] 4. It laid a solid foundation for unmanned industrial operation.

[0152] This invention constructs a closed-loop, self-decision-making intelligent agent, which greatly reduces the reliance on human experience and lays a solid foundation for realizing the industrialization and unmanned operation of waste battery recycling.

[0153] 5. Demonstrates broad applicability and excellent robustness.

[0154] The value of this invention is not limited to ideal working conditions. Its wide applicability and excellent robustness ensure that the system can effectively handle a variety of battery materials and metal raw materials. Through the built-in robust control mechanism, it can easily cope with extreme industrial working conditions such as raw material fluctuations, sensor failures, and actuator abnormalities, providing a solid guarantee for the high reliability and stable operation of the system in real and complex environments.

Claims

1. A high-efficiency metal ion leaching system based on multi-parameter dynamic coordination and digital twin, used to construct a physical-virtual-algorithm closed-loop self-evolving architecture to achieve data-driven closed-loop optimization control, characterized in that, include: The online spectral monitoring module is used to collect process parameter information of the physical leaching reaction system in real time during the leaching process. The process parameter information includes metal ion concentration data, pH value and conductivity, so as to comprehensively reflect the changes in the chemical environment during the leaching process. The leaching process digital twin module is communicatively connected to the online spectral monitoring module and is used for data information sent by the online spectral monitoring module. The high-fidelity neural network model performs simulation and deduction based on the acquired data information to predict and generate predictive control commands containing future parameter adjustment strategies. The adaptive adjustment algorithm module based on deep reinforcement learning is communicatively connected to the leaching process digital twin module and the multi-parameter collaborative controller. It is used to receive the predictive control commands issued by the leaching process digital twin module and convert them into the current weight adjustment value. The multi-parameter collaborative controller is communicatively connected to the adaptive adjustment algorithm module and the leaching process digital twin module. Its core is a collaborative optimization model with a dynamic weight allocation mechanism, which receives the current weight adjustment value, updates the weight of the collaborative optimization model in real time according to the current weight adjustment value, and outputs the final process parameter control command. The intelligent energy management module is communicatively connected to the multi-parameter collaborative controller and the leaching process digital twin module. It is used to evaluate energy efficiency and feed the evaluation results back to the leaching process digital twin module for iteration of the high-fidelity neural network model.

2. The high-efficiency metal ion leaching system based on multi-parameter dynamic coordination and digital twin as described in claim 1, characterized in that, The dynamic weight allocation mechanism is a real-time weight update mechanism driven by predictive control commands from the digital twin module of the leaching process, and its mathematical model is expressed as follows: Where: η_total is the total leaching efficiency, which is the weighted sum of the contribution values ​​of each process parameter, representing the overall optimization effect of the system; w_T(t) is a dynamic weighting factor for the reaction temperature T. It is a function of time t and is not a fixed value. It is used to represent the weight of the temperature on the overall optimization at the current moment. η_T is the contribution of reaction temperature T to the overall efficiency; w_t(t) is the dynamic weighting factor of reaction time t. It is a function of time t and is used to represent the weight of reaction time in the overall optimization at the current moment. η_t is the contribution of reaction time t to the overall efficiency; w_S / L(t) is the dynamic weighting factor of the solid-liquid ratio S / L. It is a function of time t and is used to represent the weight of the solid-liquid ratio in the overall optimization at the current moment. η_S / L is the contribution of the solid-liquid ratio S / L to the overall efficiency; w_ω(t) is the dynamic weighting factor of the stirring speed ω, which is a function of time t and is used to represent the weight of the stirring speed on the overall optimization at the current moment. η_ω is the contribution of stirring speed ω to the overall efficiency; In addition, the aforementioned weighting factor w_i(t) is not a fixed value. It is dynamically calculated by an adaptive adjustment algorithm based on the predictive instructions of the digital twin system, enabling the optimization focus to shift in real time based on future trends.

3. The high-efficiency metal ion leaching system based on multi-parameter dynamic coordination and digital twin as described in claim 1, characterized in that, The "physical-virtual-algorithm closed-loop self-evolutionary architecture" is specifically manifested as follows: the high-fidelity neural network model is calibrated in real time based on the real-time acquired operating data; the accuracy of the digital twin model improves the training effect of the reinforcement learning algorithm; and the optimized control strategy makes the physical leaching reaction system run more stably and generate higher quality data. The three form a positive feedback closed loop, which enables the overall system performance to continuously improve without human intervention. The operating data includes metal ion concentration, temperature, time, solid-liquid ratio, stirring speed, pH value, conductivity, and power.

4. The high-efficiency metal ion leaching system based on multi-parameter dynamic coordination and digital twin as described in claim 1, characterized in that, The multi-parameter collaborative controller, digital twin system, and intelligent energy management system constitute a closed-loop, self-evolving optimization architecture: the digital twin system generates predictive optimization strategies, the multi-parameter collaborative controller executes short-term precise control, and the intelligent energy management system evaluates long-term energy efficiency, with the evaluation results fed back to the digital twin system for model correction and iteration. This closed-loop, self-evolving optimization architecture enables the system to autonomously optimize its internal digital twin model and control strategies through continuous "physical-virtual-physical" data interaction without human intervention, thereby achieving continuous improvement in processing efficiency and energy efficiency.

5. The high-efficiency metal ion leaching system based on multi-parameter dynamic coordination and digital twin as described in claim 1, characterized in that, It also includes a robust control module, which acquires real-time sensor data and identifies data anomalies by comparing the confidence intervals of the real-time sensor data with the predicted values ​​in the leaching process digital twin module. The real-time sensor data includes metal ion concentration, temperature, pressure, and pH value. The predicted value refers to the estimated value of the system state at the current or future moment, calculated by the high-fidelity neural network model of the leaching process digital twin module based on real-time data. When real-time sensor data is outside the confidence interval, it is judged as abnormal. The abnormal data point is temporarily ignored, and short-term maintenance control is performed based on the previous valid data and digital twin prediction, while triggering an alarm.

6. The high-efficiency metal ion leaching system based on multi-parameter dynamic coordination and digital twin as described in claim 5, characterized in that, The robust control module can automatically trigger a safety shutdown procedure when it detects that the process parameters exceed the safe operating window. The safety shutdown procedure includes cutting off the heating source, stopping stirring, and recording a fault log. The process parameters include temperature and pressure.

7. The high-efficiency metal ion leaching system based on multi-parameter dynamic coordination and digital twin as described in claim 1, characterized in that, The adaptive adjustment algorithm module has a response time of less than 1 second for dynamically adjusting weight values.

8. The high-efficiency metal ion leaching system based on multi-parameter dynamic coordination and digital twin as described in claim 1, characterized in that, The method for constructing the digital twin module of the leaching process includes: training a high-fidelity neural network model using historical experimental data and physicochemical mechanisms; receiving data from an online spectral monitoring module and a multi-parameter collaborative controller in real time to maintain synchronization with the physical leaching reaction system; rapidly simulating the impact of different parameter combinations on the final leaching efficiency and total energy consumption in a virtual environment to predict and generate the optimal future parameter adjustment strategy; wherein the prediction time window is 0.5-2 hours in the future; the accuracy standard for prediction and extrapolation is: the average absolute error between the predicted value and the actual measured value of metal ion concentration in the next hour is less than 5%.

9. The high-efficiency metal ion leaching system based on multi-parameter dynamic coordination and digital twin as described in claim 1, characterized in that, The intelligent energy management module employs a combined optimization strategy of genetic algorithm and particle swarm optimization algorithm, using the long-term energy consumption prediction provided by the digital twin system as the objective function for optimization.

10. A leaching control method for a high-efficiency metal ion leaching system based on multi-parameter dynamic coordination and digital twin, characterized in that, Includes the following steps: Step S1: The online spectral monitoring module collects metal ion concentration data in real time during the leaching process; Step S2: The digital twin module of the leaching process performs simulation and deduction based on the acquired metal ion concentration data to predict and generate future parameter adjustment strategies, and sends predictive control commands with future parameter adjustment strategies. Step S3: The adaptive adjustment algorithm module calculates the current weight adjustment value of the collaborative optimization model based on the future parameter adjustment strategy in the predictive control command; Step S4: The multi-parameter collaborative controller receives the current weight adjustment value, updates the dynamic weight allocation mechanism of its internal collaborative optimization model in real time, calculates the final process parameters, and outputs the control command with the final process parameters to the physical leaching reaction system. Step S5: The intelligent energy management system evaluates energy efficiency in real time and feeds the evaluation results back to the leaching process digital twin module for the next round of model iteration and control.

Images