A method and apparatus for controlling wet-process phosphoric acid tandem extraction

By using a time-varying state-space prediction model in the wet phosphoric acid tandem extraction process, the detection data of the extraction tank is updated and corrected, and the operation of the flow regulation equipment is optimized, thus solving the problems of extraction rate fluctuation and energy consumption and achieving higher control accuracy and stability.

CN123105594APending Publication Date: 2026-11-17WUHAN INST OF TECH
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202611289074.0
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2026-08-24
Publication Date
2026-11-17

AI Technical Summary

Technical Problem

In the wet-process phosphoric acid tandem extraction process, there is strong material and energy coupling between each extraction tank. Traditional control methods cannot effectively handle the problems of large time delay and strong coupling, resulting in large fluctuations in extraction rate, deterioration of phosphogypsum crystal morphology, increased phosphorus encapsulation, and increased system energy consumption.

Method used

A time-varying state-space prediction model is adopted. By acquiring the current reaction process detection data of each extraction tank, the model is updated and the predicted output value is corrected. The constraint objects and control increment sequences are determined, and the flow regulation equipment is adjusted to optimize the extraction process.

Benefits of technology

It improves the control precision of the extraction process, reduces the fluctuation of the extraction rate, enhances the ability to predict and control fluctuations in upstream operating conditions, and optimizes the operating conditions of the extraction equipment.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN123105594A_ABST
    Figure CN123105594A_ABST
Patent Text Reader

Abstract

This invention relates to the field of phosphoric acid manufacturing technology, and in particular to a method and apparatus for controlling the series extraction of wet phosphoric acid. The invention updates a pre-built time-varying state prediction model using the reaction process detection data of the extraction tank at the current moment, so that the updated target time-varying state space prediction model can adapt to the current operating conditions of the extraction tank. Then, based on the difference between the expected reaction process detection data predicted in the previous control cycle and the current reaction process detection data, the output prediction value of the target time-varying state space prediction model is corrected. This improves the extraction tank's ability to coordinate prediction and control of upstream operating condition fluctuations during the extraction process. Furthermore, the corrected prediction output value determines the constraint objects and control increment sequences for each extraction tank, enabling targeted operational constraints on each extraction tank, adjusting the operating conditions of the wet phosphoric acid series extraction equipment, improving the control accuracy of the extraction process, and reducing extraction rate fluctuations.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] This invention relates to the field of phosphoric acid manufacturing technology, and in particular to a wet-process phosphoric acid tandem extraction control method and apparatus. Background Technology

[0002] In the wet process of phosphoric acid production, the reaction between phosphate rock and sulfuric acid is usually carried out in a multi-stage series extraction tank. This is a typical complex nonlinear series dynamic system. The output of the previous stage reaction tank is directly used as the input of the next stage reaction tank, resulting in strong material and energy coupling between the various reaction tanks.

[0003] Currently, the control of wet-process phosphoric acid series extraction processes mostly adopts traditional centralized control or single-loop PID control. However, due to the difficulty in establishing an accurate global multi-tank dynamic mechanism model and the dynamic changes in the reaction state within each tank over time, traditional methods cannot effectively handle large time delays and strong coupling problems. When the operating conditions of the upstream tank fluctuate (such as changes in phosphate rock grade or acid addition), it can easily cause drastic fluctuations in the temperature and supersaturation of the downstream tank, which in turn leads to the deterioration of phosphogypsum crystal morphology, increased phosphorus encapsulation, decreased phosphate rock extraction rate, and increased system energy consumption, resulting in low control accuracy and large fluctuations in the extraction rate.

[0004] The above content is only used to help understand the technical solution of the present invention and does not represent an admission that the above content is prior art. Summary of the Invention

[0005] The main objective of this invention is to provide a wet-process phosphoric acid series extraction control method and apparatus, which aims to solve the technical problems in the prior art where there is strong material and energy coupling between extraction tanks during the phosphoric acid extraction process, resulting in low control precision and large fluctuations in extraction rate in the traditional wet-process phosphoric acid extraction process.

[0006] To achieve the above objectives, the present invention provides a wet-process phosphoric acid series extraction control method, which is applied to a wet-process phosphoric acid series extraction device, wherein the wet-process phosphoric acid series extraction device includes at least a plurality of extraction tanks and flow regulating devices connected in series. The method includes the following steps: Acquire the current reaction process detection data in each extraction tank; The time-varying state space prediction model is updated based on the current reaction process detection data to obtain the target time-varying state space prediction model. The time-varying state space prediction model is a data prediction model pre-constructed for the wet phosphoric acid tandem extraction equipment, used to predict the reaction process detection data of each extraction tank in at least one future control cycle. Based on the difference between the expected reaction process detection data and the current reaction process detection data, the prediction output value of the target time-varying state space prediction model is corrected, wherein the expected reaction process detection data is the reaction process detection data of the time-varying state space prediction model in the previous period for the current period; The constraint objects and control increment sequences for each extraction cell are determined based on the corrected predicted output values. The flow regulation device is controlled based on the constraint object and the control increment sequence to adjust the operating conditions of the wet phosphoric acid tandem extraction device.

[0007] Optionally, before updating the time-varying state-space prediction model based on the current reaction process detection data, the method further includes: Determine the current state matching vector corresponding to the current reaction process detection data; Based on the current state matching vector, select multiple nearest neighbor state matching vectors from the historical database; Construct a local regression parameter matrix based on the multiple nearest neighbor state matching vectors; Extract the controlled output regression coefficients, control input coefficients, ingress condition coefficients, and bias terms from the regression parameter matrix, respectively. A time-varying state-space prediction model for each extraction tank is constructed based on the controlled output regression coefficient, control input coefficient, inlet condition coefficient, and bias term.

[0008] Optionally, the step of selecting multiple nearest neighbor state matching vectors from the historical database based on the current state matching vector includes: Multiple historical state matching vectors are determined from the historical database based on the variable order and historical order of the current state matching vector. Normalize the current state matching vector and each historical state matching vector respectively; Calculate the weighted Euclidean distance between the normalized current state matching vector and the normalized historical state matching vectors; Multiple nearest neighbor state matching vectors are selected based on the weighted Euclidean distance. The nearest neighbor state matching vectors are multiple historical state matching vectors sorted by weighted Euclidean distance from smallest to largest. The formula for calculating the weighted Euclidean distance is: in, Nearest neighbor state matching vector The dimensional variables, and The first in the historical database Minimum and maximum values ​​of a dimensional variable. This is the normalized nearest neighbor state matching vector. For the first The variable weight matrix corresponding to the extraction tank. Let i be the normalized matching vector of the i-th nearest neighbor sample in the history database. is the weighted distance between the current working condition and the i-th historical sample.

[0009] Optionally, constructing a local regression parameter matrix based on the plurality of nearest neighbor state matching vectors includes: Calculate the normalized Gaussian weights of the matching vectors of each nearest neighbor state; Construct regression feature vectors corresponding to each nearest neighbor state matching vector based on the normalized Gaussian weights; A local regression sample matrix, a local output sample matrix, and a diagonal weight matrix are generated based on the regression feature vectors. The local regression sample matrix is ​​obtained by arranging the regression feature vectors in rows. The local output sample matrix is ​​obtained by arranging the controlled output vectors of each nearest neighbor state matching vector in rows. The diagonal weight matrix is ​​constructed based on the normalized Gaussian weights of each nearest neighbor state matching vector. A local regression parameter matrix is ​​constructed based on the local regression sample matrix, the local output sample matrix, and the diagonal weight matrix. The formula for calculating the normalized Gaussian weights is as follows: in, This is the lower bound of the Gaussian kernel width, and , The adaptive Gaussian kernel width parameter at the current moment; For the first Normalized Gaussian weights for the nearest neighbor samples, and satisfying , To traverse the entire nearest neighbor sample set The distance to all samples q in the virtual traversal, where q is the index of the virtual traversal. It is the weighted distance between the current working condition and the i-th nearest neighbor sample.

[0010] Optionally, constructing the local regression sample matrix based on the local regression sample matrix, the local output sample matrix, and the diagonal weight matrix includes: in, Describing the Frobenius norm, For the first The local regression parameter matrix of the extraction tank under current operating conditions. For local regression sample matrix, For local output sample matrix, It is a diagonal weight matrix.

[0011] Optionally, the step of constructing a time-varying state-space prediction model for each stage of the extraction tank based on the controlled output regression coefficients, control input coefficients, inlet condition coefficients, and bias terms includes: Based on the current reaction process detection data, construct a control input vector, a controlled output vector, an inlet condition vector, and an auxiliary state vector; A current state matching vector is generated based on the control input vector, the controlled output vector, the entry condition vector, and the auxiliary state vector; An augmented state vector is constructed based on the control input vector, the controlled output vector, the entry condition vector, the auxiliary state vector, and historical vectors in the historical database. Extract the output selection matrix corresponding to the controlled output vector from the augmented state vector; A time-varying state-space prediction model for each extraction tank is constructed based on the augmented state vector, output selection matrix, controlled output regression coefficient, control input coefficient, inlet condition coefficient, and bias term. The formula for the time-varying state-space prediction model is: in, This is an augmented state vector, which represents the current and historical operating conditions. To control the time-varying state transition matrix corresponding to the input coefficients, The control matrix corresponding to the controlled output coefficients. This is the entry condition matrix corresponding to the entry condition coefficients. This is the bias vector corresponding to the bias term. To control the input vector, This is the entry condition vector. Predict the first time at time t The state of the extraction tank at time t+1. Choose a matrix for the output. For the time-varying state-space prediction model for the first Model prediction output for the multistage extraction cell.

[0012] Optionally, the formula for correcting the predicted output value of the target time-varying state-space prediction model includes: in, The detection data is used to predict the expected reaction process of the current cycle based on the data from the previous cycle. This is the current reaction process detection data. and , , represent the predicted output values ​​before and after correction, respectively, where P is the prediction time domain.

[0013] Optionally, determining the constraint object and control increment sequence for each extraction cell based on the corrected predicted output value includes: Determine the order of the extraction tanks along the slurry flow direction; The output deviation weight matrix and the control increment matrix are set according to the order of priority. The output deviation weight matrix is ​​used to determine the constraint objects of each extraction tank, and the control increment matrix is ​​used to determine the adjustment range of the control increment sequence. Construct process target vectors for the constraint objects of each extraction tank, wherein the process target vectors characterize the free sulfuric acid concentration and slurry temperature of the extraction tank; Based on the corrected predicted output value, the deviation weight matrix, the control increment matrix, and the process target vector, control increment optimization is performed within a preset constraint range to obtain a control increment sequence. The preset constraint range includes the control input amplitude range and the control increment amplitude range. The control input amplitude range is used to limit the maximum and minimum values ​​of the control input, and the control increment amplitude range is used to limit the maximum and minimum values ​​of the control increment. The formula for controlling incremental optimization is as follows: in, For the first The control increment sequence of the extraction cell in the control time domain. For the first The cost function for incremental optimization of the control in the extraction cell. This is the corrected predicted output value. To control the increment, To control the input amplitude, Let be the process objective vectors for free sulfuric acid concentration and slurry temperature. and These are the output deviation weight matrix and the control increment weight matrix, respectively. and These are the prediction time domain and the control time domain, respectively. and To control the minimum and maximum values ​​of the input amplitude, and To control the minimum and maximum values ​​of the increment amplitude.

[0014] Optionally, the constrained objects include at least the free sulfuric acid concentration and the slurry temperature, the flow regulating device includes at least a sulfuric acid flow regulating mechanism and a cooling medium flow regulating mechanism, the acid adding valve is used to adjust the free sulfuric acid concentration in the extraction tank, and the cooling fan is used to adjust the slurry temperature in the extraction tank; The control of the flow regulation device based on the constraint object and the control increment sequence includes: When the constraint object is the concentration of free sulfuric acid, a target control command is generated according to the control increment sequence, and the opening degree of the sulfuric acid flow regulating mechanism is adjusted based on the target control command; When the constraint object is the slurry temperature, a target control command is generated according to the control increment sequence, and the rotation speed of the cooling medium flow regulating mechanism is adjusted based on the target control command.

[0015] Furthermore, to achieve the above objectives, the present invention also proposes a wet-process phosphoric acid tandem extraction control device, the wet-process phosphoric acid tandem extraction control device comprising: The data acquisition module is used to acquire the current reaction process detection data in each extraction tank; The model update module is used to update the time-varying state space prediction model based on the current reaction process detection data to obtain the target time-varying state space prediction model. The time-varying state space prediction model is a data prediction model pre-constructed for the wet phosphoric acid tandem extraction equipment, used to predict the reaction process detection data of each extraction tank in at least one future control cycle. The output correction module is used to correct the prediction output value of the target time-varying state space prediction model based on the difference between the expected reaction process detection data and the current reaction process detection data. The expected reaction process detection data is the reaction process detection data predicted by the time-varying state space prediction model in the previous period for the current period. The incremental determination module is used to determine the constraint objects and control incremental sequences for each extraction cell based on the corrected predicted output values. The operation control module is used to control the flow regulation device based on the constraint object and the control increment sequence, so as to adjust the operating conditions of the wet phosphoric acid tandem extraction device.

[0016] Furthermore, to achieve the above objectives, the present invention also proposes a wet-process phosphoric acid tandem extraction control device, the wet-process phosphoric acid tandem extraction control device comprising: a memory, a processor, and a wet-process phosphoric acid tandem extraction control program stored in the memory and executable on the processor, the wet-process phosphoric acid tandem extraction control program being configured to implement the steps of the wet-process phosphoric acid tandem extraction control method as described above.

[0017] Furthermore, to achieve the above objectives, the present invention also proposes a storage medium storing a wet-process phosphoric acid tandem extraction control program, which, when executed by a processor, implements the steps of the wet-process phosphoric acid tandem extraction control method as described above.

[0018] This invention provides a method for controlling the tandem extraction of wet-process phosphoric acid. The method includes acquiring current reaction process detection data in each extraction tank; updating a time-varying state-space prediction model based on the current reaction process detection data to obtain a target time-varying state-space prediction model; correcting the predicted output value of the target time-varying state-space prediction model based on the difference between the desired reaction process detection data and the current reaction process detection data; determining the constraint objects and control increment sequences for each extraction tank based on the corrected predicted output values; and controlling the flow regulation device based on the constraint objects and the control increment sequences to adjust the operating conditions of the tandem extraction of wet-process phosphoric acid. Compared with existing technologies, this invention controls the flow rate of the extraction tank at the current moment... The pre-built time-varying state prediction model is updated based on the process detection data so that the updated target time-varying state space prediction model can adapt to the current operating conditions of the current extraction tank. Then, the output prediction value of the target time-varying state space prediction model is corrected based on the difference between the expected reaction process detection data predicted in the previous control cycle and the current reaction process detection data. This improves the collaborative prediction and control capability of the extraction tank against upstream operating condition fluctuations during the extraction process. The constraint objects and control increment sequences of each extraction tank are determined through the corrected prediction output values, so as to implement targeted operating constraints on each extraction tank, adjust the operating conditions of the wet phosphoric acid series extraction equipment, improve the control accuracy of the extraction process, and reduce extraction rate fluctuations. Attached Figure Description

[0019] Figure 1 This is a schematic flowchart of the first embodiment of the wet-process phosphoric acid tandem extraction control method of the present invention; Figure 2 This is a schematic diagram of the structure of a wet phosphoric acid multi-tank series extraction device according to an embodiment of the wet phosphoric acid series extraction control method of the present invention; Figure 3 This is a schematic diagram of the collaborative prediction architecture of a wet phosphoric acid multi-tank tandem extraction process according to an embodiment of the wet phosphoric acid tandem extraction control method of the present invention. Figure 4 This is a schematic flowchart of the second embodiment of the wet-process phosphoric acid tandem extraction control method of the present invention; Figure 5 This is a comparison diagram of the control effects of an embodiment of the wet-process phosphoric acid tandem extraction control method of the present invention; Figure 6 This is a structural block diagram of the first embodiment of the wet-process phosphoric acid tandem extraction control device of the present invention.

[0020] The realization of the objective, functional features and advantages of the present invention will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0021] It should be understood that the specific embodiments described herein are for illustrative purposes only and are not intended to limit the scope of the invention.

[0022] This invention provides a method for controlling the tandem extraction of wet-process phosphoric acid, referring to... Figure 1 , Figure 1 This is a schematic flowchart of the first embodiment of a wet-process phosphoric acid tandem extraction control method of the present invention.

[0023] In this embodiment, the wet-process phosphoric acid tandem extraction control method includes the following steps: Step S10: Obtain the current reaction process detection data in each extraction tank.

[0024] It should be noted that the execution subject of this embodiment can be a computer, controller, or cloud server with data processing and data transmission functions, or other devices that can achieve the same or similar functions. For ease of explanation, the controller of a wet phosphoric acid tandem extraction device will be used as an example in this embodiment and subsequent embodiments.

[0025] The wet-process phosphoric acid series extraction equipment includes at least multiple extraction tanks connected in series and a flow regulating device. The extraction tanks are unidirectionally connected in sequence along the slurry flow direction, which can realize the step-by-step extraction of phosphoric acid. The flow regulating device can adjust the free sulfuric acid concentration and slurry temperature of each extraction tank by adjusting the flow rate or operating status of different devices or components. In this embodiment and subsequent embodiments, the flow regulating device includes at least an acid adding valve and a cooling fan. The acid adding valve is used to adjust the free sulfuric acid concentration in the extraction tank, and the cooling fan is used to adjust the slurry temperature in the extraction tank. This embodiment does not impose specific limitations on this.

[0026] The reaction process monitoring data includes, but is not limited to, the amount of sulfuric acid added, the flow rate of the cooling medium, the free sulfuric acid concentration and temperature of the slurry, the inlet slurry flow rate, the inlet free sulfuric acid concentration and temperature of the inlet slurry, and the slurry level, etc., which are reaction parameters related to phosphoric acid extraction in each extraction tank. Specifically, each pipeline in each extraction tank is equipped with a temperature detection device, a liquid level detection device, and a free sulfuric acid concentration detection device, and at least one of a sulfuric acid flow rate regulation mechanism and a cooling medium flow rate regulation mechanism is configured to realize data acquisition and subsequent control and adjustment actions.

[0027] Each detection device and control actuator is connected to the DCS and its corresponding control agent. Process detection data is stored in a historical database. Adjacent control agents transmit future outlet slurry flow rate, free sulfuric acid concentration and slurry temperature prediction information along the slurry flow direction.

[0028] For ease of understanding, please refer to Figure 2 In this embodiment and subsequent embodiments, four continuously connected extraction tanks are used as an example for illustration. The wet phosphoric acid series extraction equipment includes extraction tank No. 1, extraction tank No. 2, extraction tank No. 3 and extraction tank No. 4. Extraction tank No. 1 is equipped with a slurry inlet and a sulfuric acid inlet. The slurry overflow outlets of extraction tanks No. 1 to No. 3 are respectively connected to the slurry inlet of the next extraction tank through a series conveying pipeline. Extraction tank No. 4 is equipped with a tail tank slurry outlet, so that the reaction slurry flows through extraction tanks No. 1 to No. 4 in sequence to realize phosphoric acid extraction.

[0029] It should be noted that, in order to avoid drastic changes in the reaction parameters of the reaction solution in the extraction tank due to unforeseen circumstances, this embodiment will pre-construct a time-varying state space prediction model to predict the reaction process detection data of the reaction solution in each extraction tank in the current control cycle in the previous control cycle. By controlling in advance, large fluctuations in the reaction solution in the extraction tank can be avoided. On this basis, fine-tuning is then performed by combining the actual reaction process detection data of the current control cycle, thereby maximizing the control accuracy of the extraction process and reducing the fluctuation of the extraction rate.

[0030] Among them, the time-varying state space prediction model is associated with each extraction cell of the wet phosphoric acid series extraction equipment. It is used to predict the reaction process detection data of each extraction cell in at least one control cycle in the future. It can be used to predict the free sulfuric acid concentration and slurry temperature of each extraction cell in the next adjacent control cycle.

[0031] Step S20: Update the time-varying state space prediction model based on the current reaction process detection data to obtain the target time-varying state space prediction model.

[0032] Because there is a difference between the predicted value of the time-varying state-space prediction model and the actual reaction value of the current control cycle, the extraction fluctuates greatly during actual control. Moreover, the extraction reaction has nonlinear and time-varying characteristics, and a fixed model is difficult to accurately describe the process dynamics across the entire operating range. By using the current reaction process detection data to update the time-varying state-space prediction model in real time and performing multi-step predictions based on the updated model, it is possible to match the current operating conditions in real time, effectively reduce the prediction error caused by model mismatch, and use the more accurate prediction results as input to the local model predictive control optimization problem. This ensures that the control increment obtained by the solution can achieve good tracking of the controlled output to the process setpoint, and improves the adaptive ability of the control system to cope with disturbances such as raw material fluctuations and operating condition drift.

[0033] Specifically, the model update process relies on the dataset containing the current reaction process detection data of the current control cycle to reconstruct the time-varying state space prediction model, thereby preserving the time-domain characteristics of the current control cycle. At the same time, after the update, it can also predict the reaction process detection data of the next control cycle, so as to achieve rolling iterative optimization in the next control cycle and continuously and stably extract the volatility of the reaction.

[0034] In the specific implementation, refer to Figure 3 After the model update, the target time-varying state space prediction models corresponding to each extraction tank are respectively based on the control agent corresponding to extraction tank No. 1. Control agent corresponding to extraction tank No. 2 Control agent corresponding to extraction tank No. 3 Control agent corresponding to extraction tank No. 4 For example, firstly, Based on the current operating status of extraction cell No. 1, slurry feed conditions, and candidate control input sequences, the updated time-varying state-space prediction model is used to calculate the future... The predicted sequence of free sulfuric acid concentration and slurry temperature for step 1 is then combined with the predicted future inlet slurry flow rate, sulfuric acid addition, outlet free sulfuric acid concentration, and outlet slurry temperature for extraction tank 1. Taking into account the material transfer time lag and pipeline temperature loss between extraction tanks 1 and 2, the future inlet condition sequence for extraction tank 2 is generated and sent to [the relevant authority / organization]. .

[0035] Based on the received future inlet condition sequence, the current operating status of extraction tank No. 2, and its candidate control input sequence, calculate the future inlet condition sequence of extraction tank No. 2. The free sulfuric acid concentration and slurry temperature prediction sequences were obtained step by step, and the future inlet condition sequence for extraction tank No. 3 was generated in the same manner. Following the same method, the future inlet conditions for extraction tanks No. 1 to No. 4 were obtained by recursively extrapolating along the slurry flow direction. The collaborative prediction results of the steps.

[0036] Considering material transport time lag and pipeline temperature loss, the future inlet conditions of the next-stage extraction tank are determined based on the future outlet slurry state of the previous-stage extraction tank. Under the condition that the slurry is fully mixed and the liquid level is stable in a closed-loop manner within each extraction tank, the... The outlet free sulfuric acid concentration and outlet slurry temperature of the first-stage extraction tank are equal to the predicted values ​​of the free sulfuric acid concentration and slurry temperature at the corresponding moment, respectively. The outlet slurry flow rate is determined by the inlet slurry flow rate and the amount of sulfuric acid added in this stage. Considering the material transfer time lag between adjacent extraction tanks and the temperature loss of the series conveying pipeline, the... The future inlet condition vector of the extraction cell is: in, ; For the first The extraction tank is used to extract the first stage of the extraction process. Material transfer time delay between extraction tanks, expressed as sampling period. For the first The future inlet slurry flow rate of the primary extraction tank This is the amount of sulfuric acid to be added in the future. and The first Predicted values ​​for future free sulfuric acid concentration and slurry temperature in the primary extraction tank. To account for the temperature loss of the slurry in the series conveying pipeline, when the pipeline heat loss is negligible, let .

[0037] Input the future inlet condition vector of the 2-N stage extraction cell into the first... A time-varying state-space prediction model for the extraction tank is developed, and the model is recursively extrapolated along the slurry flow direction to obtain the future state-space prediction model for each extraction tank. The free sulfuric acid concentration and slurry temperature prediction sequence are generated step by step and stored in the database for later use in the next control cycle.

[0038] Step S30: Correct the predicted output value of the target time-varying state space prediction model based on the difference between the expected reaction process detection data and the current reaction process detection data.

[0039] It should be noted that although the time-varying state-space model has been updated, improving the overall performance of the extraction reaction, local linear approximation still has modeling errors. At the same time, the process is susceptible to short-term disturbances such as fluctuations in incoming materials, and errors accumulate in multi-step predictions. Implementing feedback correction on the prediction output in the prediction time domain can use current measured information to compensate for the prediction residual of one step, correct the entire future prediction trajectory, weaken the cumulative effect of multi-step prediction errors, and use the corrected prediction output as the input of the local cost function to improve the reliability of optimization decisions, suppress the steady-state residual error of the controlled output, and enhance the adaptability of the control system to various unmeasurable short-term disturbances.

[0040] The expected reaction process detection data is the reaction process detection data for the current period predicted by the time-varying state-space prediction model in the previous control period.

[0041] Step S40: Determine the constraint objects and control increment sequences for each extraction cell based on the corrected predicted output values.

[0042] By transmitting future inlet conditions along the slurry flow direction, the downstream control agent can obtain the impact of upstream operating condition changes on the current operating status in advance, and adjust the sulfuric acid addition and cooling medium flow rate accordingly. This results in different constraints on the extraction tanks due to the different positions of the extraction tanks in the slurry flow direction. For example, the intermediate reaction tank prioritizes controlling the slurry temperature, while the tail tank prioritizes controlling the free sulfuric acid concentration.

[0043] In this embodiment, each of the multi-stage extraction tanks has an independent control agent. Each stage only solves its own local optimization problem. When controlled independently, there may be a situation where the extraction tank at this stage is significantly adjusted, or the significant adjustment affects the extraction tank at the next stage. Therefore, this embodiment mitigates the upstream and downstream coupling interference of the multi-stage extraction tanks by setting a control increment sequence.

[0044] Specifically, the control increment sequence includes control quantity and control increment. The control quantity can be the valve opening adjustment quantity, the cooling equipment speed adjustment quantity, etc., while the control increment is the change quantity compared to the previous control cycle. The control increment sequence is optimized and solved based on the corrected predicted output. On the one hand, this enables the controller to have the ability to make forward adjustments and suppress the index deviation caused by the fluctuation of incoming materials in advance. On the other hand, control increment constraints can be directly introduced to limit the single adjustment range and prevent the actuator from violently impacting the process.

[0045] Step S50: Control the flow regulation device based on the constraint object and the control increment sequence to adjust the operating conditions of the wet phosphoric acid tandem extraction device.

[0046] The flow regulation device includes at least an acid addition valve and a cooling fan. By adaptively controlling the operation of the acid addition valve and the cooling fan, the free sulfuric acid concentration and slurry temperature of any extraction tank can be adjusted. In this embodiment and subsequent embodiments, the control output is the free sulfuric acid concentration and the slurry temperature.

[0047] This embodiment updates the pre-built time-varying state prediction model using the reaction process detection data of the extraction tank at the current moment. This allows the updated target time-varying state space prediction model to adapt to the current operating conditions of the extraction tank. Then, the output prediction value of the target time-varying state space prediction model is corrected based on the difference between the expected reaction process detection data predicted in the previous control cycle and the current reaction process detection data. This improves the extraction tank's ability to coordinate prediction and control of upstream operating condition fluctuations during the extraction process. Furthermore, the corrected prediction output value determines the constraint objects and control increment sequences for each extraction tank, enabling targeted operational constraints on each extraction tank, adjusting the operating conditions of the wet phosphoric acid tandem extraction equipment, improving the control accuracy of the extraction process, and reducing extraction rate fluctuations.

[0048] refer to Figure 4 , Figure 4This is a schematic flowchart of a second embodiment of a wet-process phosphoric acid tandem extraction control method of the present invention.

[0049] Based on the first embodiment described above, in this embodiment, before step S20, the method further includes: Step S110: Determine the current state matching vector corresponding to the current reaction process detection data.

[0050] Step S120: Select multiple nearest neighbor state matching vectors from the historical database based on the current state matching vector.

[0051] Step S130: Construct a local regression parameter matrix based on the multiple nearest neighbor state matching vectors.

[0052] Step S140: Extract the controlled output regression coefficient, control input coefficient, ingress condition coefficient, and bias term from the regression parameter matrix.

[0053] Step S150: Construct a time-varying state-space prediction model for each level of extraction tank based on the controlled output regression coefficient, control input coefficient, inlet condition coefficient, and bias term.

[0054] It should be noted that the current state matching vector integrates the key process variables of the l-th stage extraction tank to fully represent the current operating conditions. Based on the current state matching vector, the nearest neighbor samples with similar operating conditions are retrieved in the historical operating condition sample library, and the weighted distance between the current operating conditions and each historical sample is calculated. Based on the distance, an adaptive Gaussian kernel weight is constructed so that the weighted least squares identification prioritizes the use of data from samples with similar operating conditions, thereby obtaining a local regression parameter matrix that adapts to the current operating conditions and effectively copes with the strong nonlinearity and time-varying characteristics of the extraction process.

[0055] The current reaction process monitoring data includes reaction parameters related to phosphoric acid extraction in each extraction tank, such as sulfuric acid addition amount, cooling medium flow rate, free sulfuric acid concentration and slurry temperature, inlet slurry flow rate, inlet free sulfuric acid concentration and inlet slurry temperature, and slurry level. Based on the role of each parameter in the extraction tank, the current reaction process monitoring data is divided into four categories: control input, controlled output, inlet conditions, and auxiliary status. Among them, the control input is sulfuric acid addition amount and cooling medium flow rate; the controlled output is free sulfuric acid concentration and slurry temperature; the inlet conditions are inlet slurry flow rate, inlet free sulfuric acid concentration, and inlet slurry temperature; and the auxiliary status is slurry level.

[0056] Specifically, the vector representations corresponding to the control input, controlled output, entry conditions, and auxiliary states are as follows: In the formula, and The first The amount of sulfuric acid added and the flow rate of the cooling medium in the extraction tank; and The first The concentration of free sulfuric acid and the temperature of the slurry in the extraction tank; , and These are the inlet slurry flow rate, the inlet free sulfuric acid concentration, and the inlet slurry temperature, respectively. For slurry level, To control the input vector, For the controlled output vector, This is the entry condition vector. This is the auxiliary state vector.

[0057] For the first-stage extraction cell, the inlet condition vector The determination is based on the flow rate of the slurry feed, the concentration of free sulfuric acid, and the temperature detection results; for the first... Multistage extraction tank, Its entry condition vector According to the The state of the slurry at the outlet of the extraction tank is determined by the series conveying pipeline.

[0058] In addition, the current and historical values ​​of the collected control inputs, controlled outputs, entry conditions, and auxiliary states can be stored in a historical database for later retrieval.

[0059] Based on the collected data The controlled output vector, control input vector, inlet condition vector, and auxiliary state vector of the extraction cell are used to construct the current time step using process data at the current time and a preset lag time. State matching vector: In the formula, This indicates that the vectors within the parentheses will be concatenated vertically in the listed order. , , and The historical order is used for the controlled output, control input, entry condition, and auxiliary state, respectively, while the cutoff time is used for the control input. Control variables that have already been executed are used to avoid using control inputs that have not yet been solved in the current control cycle during state matching.

[0060] Historical operating data of each extraction tank is collected in advance according to a preset sampling period. The collected data is then synchronized in time, outlier removal is performed, and missing values ​​are processed to form a historical modeling database corresponding to each extraction tank. In this embodiment, the sampling period is 1 minute, and the historical modeling database contains process operating data for three consecutive months.

[0061] For the Multistage extraction cell, based on controlled output vector Control input vector , entry condition vector and auxiliary state vector Construct the state matching vector for the current moment. The controlled output vector includes the free sulfuric acid concentration and slurry temperature, the control input vector includes the sulfuric acid addition amount and the cooling medium flow rate, the inlet condition vector includes the inlet slurry flow rate, the inlet free sulfuric acid concentration and the inlet slurry temperature, and the auxiliary state vector includes the slurry level.

[0062] The control input in the state matching vector uses the historical control quantities that have been actually executed before the start of the current control cycle, and does not use the control quantities that have not yet been optimized in the current control cycle. Each historical state matching vector in the historical database is constructed according to the same variable order and historical order as the current state matching vector.

[0063] This embodiment selects historical nearest neighbor conditions by normalized weighted distance and identifies local regression parameters online based on the weights of nearest neighbor samples, enabling the prediction model to be updated in a rolling manner as the feeding status and process operating conditions change, thereby improving the model's adaptability to operating conditions.

[0064] Further, the step of selecting multiple nearest neighbor state matching vectors from the historical database based on the current state matching vector includes: Multiple historical state matching vectors are determined from the historical database based on the variable order and historical order of the current state matching vector. Normalize the current state matching vector and each historical state matching vector respectively; Calculate the weighted Euclidean distance between the normalized current state matching vector and the normalized historical state matching vectors; Multiple nearest neighbor state matching vectors are selected based on the weighted Euclidean distance. The nearest neighbor state matching vectors are multiple historical state matching vectors sorted by weighted Euclidean distance from smallest to largest.

[0065] Construct the first [vector] from the historical database according to the same variable order and historical order as the current state matching vector. Historical state matching vectors Based on the minimum and maximum values ​​of each dimension of the variables in the historical database, the current state matching vector and the historical state matching vector are normalized dimension by dimension to obtain the normalized current state matching vector. Matching vector with normalized history state ; Calculate the normalized current state matching vector Compared with the historical database Normalized historical state matching vector Weighted Euclidean distance between And select the first ones in ascending order of the weighted Euclidean distance. Each historical state matching vector serves as a similar operating condition sample for the current operating condition, forming a nearest neighbor sample index set. This is used for the subsequent calculation of normalized Gaussian weights.

[0066] The formula for calculating the weighted Euclidean distance is: in, Nearest neighbor state matching vector The dimensional variables, and The first in the historical database Minimum and maximum values ​​of a dimensional variable. This is the normalized nearest neighbor state matching vector. For the first The variable weight matrix corresponding to the extraction tank. Let i be the normalized matching vector of the i-th nearest neighbor sample in the history database. is the weighted Euclidean distance between the current working condition and the i-th historical sample.

[0067] Further, the step of constructing a local regression parameter matrix based on the plurality of nearest neighbor state matching vectors includes: Calculate the normalized Gaussian weights of the matching vectors of each nearest neighbor state; Construct regression feature vectors corresponding to each nearest neighbor state matching vector based on the normalized Gaussian weights; A local regression sample matrix, a local output sample matrix, and a diagonal weight matrix are generated based on the regression feature vectors. The local regression sample matrix is ​​obtained by arranging the regression feature vectors in rows. The local output sample matrix is ​​obtained by arranging the controlled output vectors of each nearest neighbor state matching vector in rows. The diagonal weight matrix is ​​constructed based on the normalized Gaussian weights of each nearest neighbor state matching vector. A local regression parameter matrix is ​​constructed based on the local regression sample matrix, the local output sample matrix, and the diagonal weight matrix. The formula for calculating the normalized Gaussian weights is as follows: in, This is the lower bound of the Gaussian kernel width, and , The adaptive Gaussian kernel width parameter at the current moment; For the first Normalized Gaussian weights for the nearest neighbor samples, and satisfying , To traverse the entire nearest neighbor sample set The distance to all samples q in the virtual traversal, where q is the index of the virtual traversal. It is the weighted distance between the current working condition and the i-th nearest neighbor sample.

[0068] Based on the minimum and maximum values ​​of each dimension of the variables in the historical modeling database, the current state matching vector and the historical state matching vector are normalized dimension by dimension, and the weighted Euclidean distance between them is calculated according to the above formula. Since each dimension of the variables has been dimensionless before the distance calculation, the variable weight matrix in this embodiment adopts an equal weight setting, that is, the variable weight matrix is ​​set to... ,in, It is an identity matrix.

[0069] Select the first ones in ascending order of weighted Euclidean distance. Each historical state matching vector constitutes a nearest neighbor sample set corresponding to the current operating condition. The number of nearest neighbor samples is... Based on historical validation data, the following methods are determined: Local prediction models are established within a preset candidate range using different numbers of nearest neighbor samples. The root mean square error (RMSE) of each model's single-step prediction on the historical validation data is calculated, and the sample number corresponding to the minimum error is determined as [the minimum number of samples]. .

[0070] Use the selected The median distance between the nearest neighbor samples is used as the base kernel width, and a preset positive number is used to... By limiting the minimum value of the kernel width parameter, the Gaussian kernel width is reduced when the historical samples near the working condition are densely distributed, and increased when the historical samples are sparsely distributed. Then, the normalized Gaussian weight is calculated based on the distance between each neighboring sample and the current state matching vector, so that historical samples closer to the current working condition receive a larger modeling weight.

[0071] Further, the step of constructing the local regression sample matrix based on the local regression sample matrix, the local output sample matrix, and the diagonal weight matrix includes: in, Describing the Frobenius norm, For the first The local regression parameter matrix of the extraction tank under current operating conditions. For local regression sample matrix, For local output sample matrix, It is a diagonal weight matrix.

[0072] Based on the selected nearest neighbor samples and their normalized weights, a local regression sample matrix, a local output sample matrix, and a diagonal weight matrix are constructed, respectively. Then, the weighted least squares method is used to identify the first nearest neighbor sample online. The local regression parameter matrix corresponding to the current operating conditions of the extraction cell.

[0073] Specifically, for each similar operating condition sample, a regression feature vector is constructed in the order of historical controlled output values, current and historical control input values, current and historical input conditions, and constant terms. The regression feature vectors are arranged in rows to form a local regression sample matrix. Arrange the controlled output vectors corresponding to each sample at the next time step in rows to form a local output sample matrix. Construct a diagonal weight matrix based on the normalized weights of each similar working condition sample: .

[0074] Further, the step of constructing a time-varying state-space prediction model for each stage of the extraction tank based on the controlled output regression coefficients, control input coefficients, inlet condition coefficients, and bias terms includes: Based on the current reaction process detection data, construct a control input vector, a controlled output vector, an inlet condition vector, and an auxiliary state vector; A current state matching vector is generated based on the control input vector, the controlled output vector, the entry condition vector, and the auxiliary state vector; An augmented state vector is constructed based on the control input vector, the controlled output vector, the entry condition vector, the auxiliary state vector, and historical vectors in the historical database. Extract the output selection matrix corresponding to the controlled output vector from the augmented state vector; A time-varying state-space prediction model for each extraction tank is constructed based on the augmented state vector, output selection matrix, controlled output regression coefficient, control input coefficient, inlet condition coefficient, and bias term. The formula for the time-varying state-space prediction model is: in, This is an augmented state vector, which represents the current and historical operating conditions. To control the time-varying state transition matrix corresponding to the input coefficients, The control matrix corresponding to the controlled output coefficients. This is the entry condition matrix corresponding to the entry condition coefficients. This is the bias vector corresponding to the bias term. To control the input vector, This is the entry condition vector. Predict the first time at time t The state of the extraction tank at time t+1. Choose a matrix for the output. For the time-varying state-space prediction model for the first Model prediction output for the multistage extraction cell.

[0075] At the beginning of each control cycle, state matching, nearest neighbor sample selection, sample weight calculation, and local regression parameter identification are re-performed based on the latest acquired process data to update the time-varying state-space prediction model for the current control cycle. Within the prediction time domain of the same control cycle, the model parameters are kept unchanged, and the future predictions are recursively calculated using the above formula. The predicted output at each sampling time is updated again with similar working condition samples and model parameters after new measurement data is acquired in the next control cycle.

[0076] Based on the order of controlled output, control input, inlet conditions, and constant terms in the regression feature vector, the corresponding output regression coefficients, control input coefficients, inlet condition coefficients, and bias terms are extracted from the local regression parameter matrix. A time-varying state space prediction model is then constructed based on this model. The time-varying state space prediction model uses free sulfuric acid concentration and slurry temperature as controlled outputs, sulfuric acid addition amount and cooling medium flow rate as control inputs, and inlet slurry flow rate, inlet free sulfuric acid concentration, and inlet slurry temperature as inlet conditions.

[0077] This embodiment determines the current state matching vector corresponding to the current reaction process detection data; selects multiple nearest neighbor state matching vectors from the historical database based on the current state matching vector; constructs a local regression parameter matrix based on the multiple nearest neighbor state matching vectors; extracts the controlled output regression coefficient, control input coefficient, inlet condition coefficient, and bias term from the regression parameter matrix; constructs a time-varying state space prediction model for each extraction tank based on the controlled output regression coefficient, control input coefficient, inlet condition coefficient, and bias term; and achieves multi-step rolling prediction of free sulfuric acid concentration and slurry temperature in each extraction tank by establishing the correspondence between the local regression parameters of each extraction tank and the time-varying state space prediction model.

[0078] In one feasible embodiment, the formula for correcting the predicted output value of the target time-varying state-space prediction model includes: in, The detection data is used to predict the expected reaction process of the current cycle based on the data from the previous cycle. This is the current reaction process detection data. and , , represent the predicted output values ​​before and after correction, respectively, where P is the prediction time domain.

[0079] Although the time-varying state-space model is updated in real time through local weighted identification, a mismatch still exists between the model and the actual process due to the influence of local linear approximation, unmeasurable disturbances, and coupling of multi-level extraction units. Correcting the predicted output value of the target time-varying state-space prediction model can use the measured information of the process to correct the internal state of the model, suppress model deviation from the source, alleviate the accumulation of errors in the multi-step recursive process, provide a reliable model basis for subsequent local model predictive control optimization solutions, and improve the anti-disturbance capability of the control system.

[0080] Further, determining the constraint objects and control increment sequences for each extraction cell based on the corrected predicted output values ​​includes: Determine the order of the extraction tanks along the slurry flow direction; The output deviation weight matrix and the control increment matrix are set according to the order of priority. The output deviation weight matrix is ​​used to determine the constraint objects of each extraction tank, and the control increment matrix is ​​used to determine the adjustment range of the control increment sequence. Construct process target vectors for the constraint objects of each extraction tank, wherein the process target vectors characterize the free sulfuric acid concentration and slurry temperature of the extraction tank; Based on the corrected predicted output value, the deviation weight matrix, the control increment matrix, and the process target vector, control increment optimization is performed within a preset constraint range to obtain a control increment sequence. The preset constraint range includes the control input amplitude range and the control increment amplitude range. The control input amplitude range is used to limit the maximum and minimum values ​​of the control input, and the control increment amplitude range is used to limit the maximum and minimum values ​​of the control increment.

[0081] The formula for controlling incremental optimization is as follows: in, For the first The control increment sequence of the extraction cell in the control time domain. For the first The cost function for incremental optimization of the control in the extraction cell. This is the corrected predicted output value. To control the increment, To control the input amplitude, Let be the process objective vectors for free sulfuric acid concentration and slurry temperature. and These are the output deviation weight matrix and the control increment weight matrix, respectively. and These are the prediction time domain and the control time domain, respectively. and To control the minimum and maximum values ​​of the input amplitude, and To control the minimum and maximum values ​​of the increment amplitude.

[0082] Each level of control agent sets the output deviation weight matrix. The weights corresponding to different output components are used to determine the main control targets of each extraction tank. Among them, the intermediate reaction tank prioritizes the control of slurry temperature, while the tail tank prioritizes the control of free sulfuric acid concentration.

[0083] refer to Figure 5 Adjacent control agents exchange information on future outlet slurry flow rate, free sulfuric acid concentration, and temperature prediction along the slurry flow direction. After each control agent completes the local optimization solution, it only issues the first control command in the optimal control sequence to the corresponding sulfuric acid flow regulating mechanism (e.g., acid adding valve) and cooling medium flow regulating mechanism (e.g., cooling fan). After acquiring new process detection data in the next control cycle, it uses the predicted output value of the target time-varying state space prediction model in the current control cycle to re-perform state matching, model update, series collaborative prediction, and optimization solution.

[0084] In this embodiment, each of the multi-stage extraction tanks has an independent control agent. Each stage only solves its own local optimization problem, which reduces the computational burden compared to centralized MPC and facilitates engineering implementation. At the same time, the control increment optimization can achieve high-precision steady-state tracking of free sulfuric acid concentration and slurry temperature while meeting the constraints of equipment amplitude and action increment. It also smooths the control actions, suppresses process oscillations, and significantly improves the anti-disturbance capability and adaptive control performance of the multi-stage coupled extraction system.

[0085] In one feasible embodiment, the constrained objects include at least the free sulfuric acid concentration and the slurry temperature, the flow regulating device includes at least a sulfuric acid flow regulating mechanism and a cooling medium flow regulating mechanism, the acid adding valve is used to adjust the free sulfuric acid concentration in the extraction tank, and the cooling fan is used to adjust the slurry temperature in the extraction tank; The control of the flow regulation device based on the constraint object and the control increment sequence includes: When the constraint object is the concentration of free sulfuric acid, a target control command is generated according to the control increment sequence, and the opening degree of the sulfuric acid flow regulating mechanism is adjusted based on the target control command; When the constraint object is the slurry temperature, a target control command is generated according to the control increment sequence, and the rotation speed of the cooling medium flow regulating mechanism is adjusted based on the target control command.

[0086] In the extraction tanks of each stage along the slurry flow direction, since the inlet conditions of the first-stage extraction tank are not affected by the other stages of extraction tanks, but only by the inlet slurry flow rate and the amount of sulfuric acid added, the coupling effect is relatively low. The constraint object can be either the free sulfuric acid concentration or the slurry temperature. For the intermediate stage extraction tanks, the slurry temperature is the primary constraint object, while for the last stage tail tank, the free sulfuric acid concentration is the primary constraint object, thereby improving the extraction rate and yield of wet phosphoric acid extraction.

[0087] Furthermore, the sulfuric acid flow regulating mechanism in this embodiment is not limited to an acid adding valve, but can also be a centrifugal pump or other equipment that controls the flow rate of the slurry, or other equipment or combined mechanisms that can control the concentration of free sulfuric acid in the tank. Similarly, the cooling medium flow regulating mechanism is not limited to a cooling fan, but can also be a cooling grid, an electric heating device, or other equipment or combined mechanisms that can control the temperature in the tank. This embodiment does not impose specific limitations on these aspects.

[0088] Furthermore, embodiments of the present invention also propose a storage medium storing a wet-process phosphoric acid tandem extraction control program, which, when executed by a processor, implements the steps of the wet-process phosphoric acid tandem extraction control method as described above.

[0089] Since this storage medium adopts all the technical solutions of all the above embodiments, it has at least all the beneficial effects brought about by the technical solutions of the above embodiments, which will not be repeated here.

[0090] Reference Figure 6 , Figure 6 This is a structural block diagram of the first embodiment of the wet-process phosphoric acid tandem extraction control device of the present invention.

[0091] like Figure 6 As shown, the wet-process phosphoric acid tandem extraction control device proposed in this embodiment of the invention includes: The data acquisition module 10 is used to acquire the detection data of the current reaction process in each extraction tank.

[0092] The model update module 20 is used to update the time-varying state space prediction model based on the current reaction process detection data to obtain the target time-varying state space prediction model. The time-varying state space prediction model is a data prediction model pre-constructed for the wet phosphoric acid tandem extraction equipment, used to predict the reaction process detection data of each extraction tank in at least one future control cycle.

[0093] The output correction module 30 is used to correct the prediction output value of the target time-varying state space prediction model based on the difference between the expected reaction process detection data and the current reaction process detection data. The expected reaction process detection data is the reaction process detection data predicted by the time-varying state space prediction model in the previous period for the current period.

[0094] The incremental determination module 40 is used to determine the constraint objects and control incremental sequences of each extraction cell based on the corrected predicted output values.

[0095] The operation control module 50 is used to control the flow regulation device based on the constraint object and the control increment sequence to adjust the operating conditions of the wet phosphoric acid tandem extraction device.

[0096] This embodiment updates the pre-built time-varying state prediction model using the reaction process detection data of the extraction tank at the current moment. This allows the updated target time-varying state space prediction model to adapt to the current operating conditions of the extraction tank. Then, the output prediction value of the target time-varying state space prediction model is corrected based on the difference between the expected reaction process detection data predicted in the previous control cycle and the current reaction process detection data. This improves the extraction tank's ability to coordinate prediction and control of upstream operating condition fluctuations during the extraction process. Furthermore, the corrected prediction output value determines the constraint objects and control increment sequences for each extraction tank, enabling targeted operational constraints on each extraction tank, adjusting the operating conditions of the wet phosphoric acid tandem extraction equipment, improving the control accuracy of the extraction process, and reducing extraction rate fluctuations.

[0097] It should be understood that the above are merely illustrative examples and do not constitute any limitation on the technical solutions of the present invention. In specific applications, those skilled in the art can make settings as needed, and the present invention does not impose any restrictions on this.

[0098] It should be noted that the workflow described above is merely illustrative and does not limit the scope of protection of this invention. In practical applications, those skilled in the art can select some or all of the workflow to achieve the purpose of this embodiment according to actual needs, and no restrictions are imposed here.

[0099] In addition, for technical details not described in detail in this embodiment, please refer to the wet phosphoric acid tandem extraction control method provided in any embodiment of the present invention, which will not be repeated here.

[0100] Furthermore, it should be noted that, in this document, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or system that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or system. Unless otherwise specified, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or system that includes that element.

[0101] The sequence numbers of the above embodiments of the present invention are for descriptive purposes only and do not represent the superiority or inferiority of the embodiments.

[0102] Through the above description of the embodiments, those skilled in the art can clearly understand that the methods of the above embodiments can be implemented by means of software plus necessary general-purpose hardware platforms. Of course, they can also be implemented by hardware, but in many cases the former is a better implementation method. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product is stored in a storage medium (such as read-only memory (ROM) / RAM, magnetic disk, optical disk) and includes several instructions to cause a terminal device (which may be a mobile phone, computer, server, or network device, etc.) to execute the methods described in the various embodiments of the present invention.

[0103] The above are merely preferred embodiments of the present invention and do not limit the scope of the patent. Any equivalent structural or procedural transformations made based on the description and drawings of the present invention, or direct or indirect applications in other related technical fields, are similarly included within the scope of patent protection of the present invention.

Claims

1. A method for controlling the tandem extraction of wet-process phosphoric acid, characterized in that, The wet phosphoric acid series extraction control method is applied to a wet phosphoric acid series extraction device, which includes at least a plurality of extraction tanks and flow regulating devices connected in series. The wet-process phosphoric acid tandem extraction control method includes: Acquire the current reaction process detection data in each extraction tank; The time-varying state space prediction model is updated based on the current reaction process detection data to obtain the target time-varying state space prediction model. The time-varying state space prediction model is a data prediction model pre-constructed for the wet phosphoric acid tandem extraction equipment, used to predict the reaction process detection data of each extraction tank in at least one future control cycle. Based on the difference between the expected reaction process detection data and the current reaction process detection data, the prediction output value of the target time-varying state space prediction model is corrected, wherein the expected reaction process detection data is the reaction process detection data of the time-varying state space prediction model in the previous period for the current period; The constraint objects and control increment sequences for each extraction cell are determined based on the corrected predicted output values. The flow regulation device is controlled based on the constraint object and the control increment sequence to adjust the operating conditions of the wet phosphoric acid tandem extraction device.

2. The wet-process phosphoric acid tandem extraction control method as described in claim 1, characterized in that, Before updating the time-varying state-space prediction model based on the current reaction process detection data, the method further includes: Determine the current state matching vector corresponding to the current reaction process detection data; Based on the current state matching vector, select multiple nearest neighbor state matching vectors from the historical database; Construct a local regression parameter matrix based on the multiple nearest neighbor state matching vectors; Extract the controlled output regression coefficients, control input coefficients, ingress condition coefficients, and bias terms from the regression parameter matrix, respectively. A time-varying state-space prediction model for each extraction tank is constructed based on the controlled output regression coefficient, control input coefficient, inlet condition coefficient, and bias term.

3. The wet-process phosphoric acid tandem extraction control method as described in claim 2, characterized in that, The step of selecting multiple nearest neighbor state matching vectors from the historical database based on the current state matching vector includes: Multiple historical state matching vectors are determined from the historical database based on the variable order and historical order of the current state matching vector. Normalize the current state matching vector and each historical state matching vector respectively; Calculate the weighted Euclidean distance between the normalized current state matching vector and the normalized historical state matching vectors; Multiple nearest neighbor state matching vectors are selected based on the weighted Euclidean distance. The nearest neighbor state matching vectors are multiple historical state matching vectors sorted by weighted Euclidean distance from smallest to largest. The formula for calculating the weighted Euclidean distance is: in, Nearest neighbor state matching vector The dimensional variables, and The first in the historical database Minimum and maximum values ​​of a dimensional variable. This is the normalized nearest neighbor state matching vector. For the first The variable weight matrix corresponding to the extraction tank. Let i be the normalized matching vector of the i-th nearest neighbor sample in the history database. is the weighted distance between the current working condition and the i-th historical sample.

4. The wet-process phosphoric acid tandem extraction control method as described in claim 2, characterized in that, The step of constructing a local regression parameter matrix based on the multiple nearest neighbor state matching vectors includes: Calculate the normalized weights of the matching vectors of each nearest neighbor state; Based on the normalized weights, construct the regression feature vectors corresponding to each nearest neighbor state matching vector; A local regression sample matrix, a local output sample matrix, and a diagonal weight matrix are generated based on the regression feature vectors. The local regression sample matrix is ​​obtained by arranging the regression feature vectors in rows. The local output sample matrix is ​​obtained by arranging the controlled output vectors of each nearest neighbor state matching vector in rows. The diagonal weight matrix is ​​constructed based on the normalized weights of each nearest neighbor state matching vector. A local regression parameter matrix is ​​constructed based on the local regression sample matrix, the local output sample matrix, and the diagonal weight matrix. The formula for calculating the normalized weights is as follows: in, This is the lower bound of the Gaussian kernel width, and , The adaptive Gaussian kernel width parameter at the current moment; For the first The normalized weights of the nearest neighbor samples, and satisfying , To traverse the entire nearest neighbor sample set The distance to all samples q in the virtual traversal, where q is the virtual traversal index. It is the weighted distance between the current working condition and the i-th nearest neighbor sample.

5. The wet-process phosphoric acid tandem extraction control method as described in claim 4, characterized in that, The step of constructing the local regression sample matrix based on the local regression sample matrix, the local output sample matrix, and the diagonal weight matrix includes: in, Describing the Frobenius norm, For the first The local regression parameter matrix of the extraction tank under current operating conditions. For local regression sample matrix, For local output sample matrix, It is a diagonal weight matrix.

6. The wet-process phosphoric acid tandem extraction control method as described in claim 2, characterized in that, The construction of time-varying state-space prediction models for each stage of the extraction tank based on the controlled output regression coefficients, control input coefficients, inlet condition coefficients, and bias terms includes: Based on the current reaction process detection data, construct a control input vector, a controlled output vector, an inlet condition vector, and an auxiliary state vector; A current state matching vector is generated based on the control input vector, the controlled output vector, the entry condition vector, and the auxiliary state vector; An augmented state vector is constructed based on the control input vector, the controlled output vector, the entry condition vector, the auxiliary state vector, and historical vectors in the historical database. Extract the output selection matrix corresponding to the controlled output vector from the augmented state vector; A time-varying state-space prediction model for each extraction tank is constructed based on the augmented state vector, output selection matrix, controlled output regression coefficient, control input coefficient, inlet condition coefficient, and bias term. The formula for the time-varying state-space prediction model is: in, This is an augmented state vector, which represents the current and historical operating conditions. To control the time-varying state transition matrix corresponding to the input coefficients, The control matrix corresponding to the controlled output coefficients. This is the entry condition matrix corresponding to the entry condition coefficients. This is the bias vector corresponding to the bias term. To control the input vector, This is the entry condition vector. Predict the first time at time t The state of the extraction tank at time t+1. Choose a matrix for the output. For the time-varying state-space prediction model for the first Model prediction output for the multistage extraction cell.

7. The wet-process phosphoric acid tandem extraction control method as described in claim 1, characterized in that, The formula for correcting the predicted output value of the target time-varying state-space prediction model includes: in, The detection data is used to predict the expected reaction process of the current cycle based on the data from the previous cycle. This is the current reaction process detection data. and , , represent the predicted output values ​​before and after correction, respectively, where P is the prediction time domain.

8. The wet-process phosphoric acid tandem extraction control method as described in claim 1, characterized in that, The step of determining the constraint objects and control increment sequences for each extraction cell based on the corrected predicted output values ​​includes: Determine the order of the extraction tanks along the slurry flow direction; The output deviation weight matrix and the control increment matrix are set according to the order of priority. The output deviation weight matrix is ​​used to determine the constraint objects of each extraction tank, and the control increment matrix is ​​used to determine the adjustment range of the control increment sequence. Construct process target vectors for the constraint objects of each extraction tank, wherein the process target vectors characterize the free sulfuric acid concentration and slurry temperature of the extraction tank; Based on the corrected predicted output value, the deviation weight matrix, the control increment matrix, and the process target vector, control increment optimization is performed within a preset constraint range to obtain a control increment sequence. The preset constraint range includes the control input amplitude range and the control increment amplitude range. The control input amplitude range is used to limit the maximum and minimum values ​​of the control input, and the control increment amplitude range is used to limit the maximum and minimum values ​​of the control increment. The formula for controlling incremental optimization is as follows: in, For the first The control increment sequence of the extraction cell in the control time domain. For the first The cost function for incremental optimization of the control in the multistage extraction cell. This is the corrected predicted output value. To control the increment, To control the input amplitude, Let be the process objective vectors for free sulfuric acid concentration and slurry temperature. and These are the output deviation weight matrix and the control increment weight matrix, respectively. and These are the prediction time domain and the control time domain, respectively. and To control the minimum and maximum values ​​of the input amplitude, and To control the minimum and maximum values ​​of the increment magnitude.

9. The wet-process phosphoric acid tandem extraction control method as described in claim 1, characterized in that, The constraints include at least the free sulfuric acid concentration and the slurry temperature; the flow regulation device includes at least a sulfuric acid flow regulation mechanism and a cooling medium flow regulation mechanism; the acid addition valve is used to adjust the free sulfuric acid concentration in the extraction tank; and the cooling fan is used to adjust the slurry temperature in the extraction tank. The control of the flow regulation device based on the constraint object and the control increment sequence includes: When the constraint object is the concentration of free sulfuric acid, a target control command is generated according to the control increment sequence, and the opening degree of the sulfuric acid flow regulating mechanism is adjusted based on the target control command; When the constraint object is the slurry temperature, a target control command is generated according to the control increment sequence, and the rotation speed of the cooling medium flow regulating mechanism is adjusted based on the target control command.

10. A wet-process phosphoric acid tandem extraction control device, characterized in that, The wet-process phosphoric acid tandem extraction control device includes: The data acquisition module is used to acquire the current reaction process detection data in each extraction tank; The model update module is used to update the time-varying state space prediction model based on the current reaction process detection data to obtain the target time-varying state space prediction model. The time-varying state space prediction model is a data prediction model pre-constructed for the wet phosphoric acid tandem extraction equipment, used to predict the reaction process detection data of each extraction tank in at least one future control cycle. The output correction module is used to correct the prediction output value of the target time-varying state space prediction model based on the difference between the expected reaction process detection data and the current reaction process detection data. The expected reaction process detection data is the reaction process detection data predicted by the time-varying state space prediction model in the previous period for the current period. The incremental determination module is used to determine the constraint objects and control incremental sequences for each extraction cell based on the corrected predicted output values. The operation control module is used to control the flow regulation device based on the constraint object and the control increment sequence, so as to adjust the operating conditions of the wet phosphoric acid tandem extraction device.