A model predictive control method, device, processor, and storage medium for temperature sensing plate in a distillation column based on LSTM.

Abstract

This invention relates to a model predictive control method for sensitive plate temperature in a distillation column based on LSTM, comprising: collecting historical data of multiple variables during the distillation column production process to obtain raw process data arranged along a time axis; calculating the maximum information coefficient (MIC) of multiple variables at different delay times; using the delay time corresponding to the auxiliary variable reaching its maximum MIC value as the optimal delay time between the auxiliary variable and the sensitive plate temperature; performing time-delay reconstruction on the raw data based on the optimal delay time to obtain a new dataset; using the time-delay reconstructed dataset as model training samples to build and train an LSTM model; using the trained LSTM model as the real-time prediction model for model predictive control (MPC), designing a sensitive plate temperature controller based on this model, and solving a nonlinear constraint optimization problem by establishing an objective optimization function to obtain the model predictive optimal control quantity. This invention also relates to a corresponding device, processor, and computer-readable storage medium thereof.

Description

Technical Field

[0001] This invention relates to the field of distillation column temperature control, specifically to a model predictive control method, device, processor, and computer-readable storage medium for the temperature of the sensitive plate of a distillation column based on LSTM. Background Technology

[0002] Distillation is a common and important separation operation in chemical production, widely used in petrochemical, pharmaceutical, and food industries. The operating conditions of the distillation column directly affect the purity and yield of the product, with temperature control being a key parameter. Traditional temperature control methods for distillation columns typically employ PID control; however, due to the highly nonlinear, strongly coupled, and large time lag characteristics of the distillation process, PID control often struggles to achieve ideal control results under complex operating conditions.

[0003] Traditional MPC controllers mostly rely on complex physical models. In actual production, due to the dynamic properties of the fluid in the distillation column, there are very complex nonlinear relationships between different variables, which leads to insufficient prediction accuracy of the prediction model. As a result, the actual effect of traditional MPC controllers cannot meet the required accuracy and effect.

[0004] Therefore, it is necessary to propose a solution to overcome the shortcomings of existing technologies. Summary of the Invention

[0005] The purpose of this invention is to overcome the shortcomings of the prior art and provide a model predictive control method, device, processor and computer-readable storage medium for the temperature of the sensitive plate of a distillation column based on LSTM.

[0006] To achieve the above objectives, the present invention provides a model predictive control method, apparatus, processor, and computer-readable storage medium for the temperature of a sensitive plate in a distillation column based on LSTM, as follows:

[0007] The main feature of this LSTM-based model predictive control method for the temperature of the sensitive plate in a distillation column is that the method includes the following steps:

[0008] (1) Collect historical data of multiple variables during the production process of the distillation column in the factory to obtain the original process data arranged in time axis;

[0009] (2) Calculate the maximum information coefficient MIC of multiple variables under different delay times, use the MIC value as an evaluation index of the correlation between auxiliary variables and the temperature of the sensitive plate of the distillation column, and determine the auxiliary variables required to predict the temperature of the sensitive plate through variable screening.

[0010] (3) The delay time corresponding to the maximum MIC value of the auxiliary variable is taken as the optimal delay time between the auxiliary variable and the temperature of the sensitive plate;

[0011] (4) Based on the optimal delay time, the original data is reconstructed to obtain a new dataset;

[0012] (5) Use the dataset obtained after time delay reconstruction as the model training sample to build and train the LSTM model;

[0013] (6) The trained LSTM model is used as the real-time prediction model of the model predictive control (MPC) to design a sensitive plate temperature controller.

[0014] (7) Establish the objective optimization function based on the predicted output of the LSTM model and the reference value in the prediction time domain, and obtain the optimal control quantity predicted by the model by solving the nonlinear constraint optimization problem.

[0015] Preferably, step (1) specifically comprises:

[0016] Historical data on the production process of the distillation column were collected from the factory, including multiple variables such as the inlet flow rate, outlet flow rate, reboiler heat carrier flow rate, column top reflux flow rate, and distillation column sensitive plate temperature. The data were sampled according to the sampling period to obtain the raw process data arranged on the time axis.

[0017] Preferably, step (2) specifically includes the following steps:

[0018] (2.1) Calculate the maximum information coefficient (MIC) between the feature variables and the target variable in the collected historical production process data under different delay times;

[0019] (2.2) Take the MIC between each feature variable and the target variable as the correlation index, remove feature variables with MIC below the set threshold, perform preliminary variable screening on multiple feature variables in the production data, and obtain auxiliary variables for predicting the temperature of the sensitive plate of the distillation column.

[0020] Preferably, step (2.1) specifically includes the following steps:

[0021] Suppose we have two variables, X and Y, containing n measurements.

[0022] (2.1.1) Divide variables X and Y into several intervals and form a two-dimensional grid to achieve data discretization;

[0023] (2.1.2) For each grid, count the number of data points in each cell to form a two-dimensional frequency table and calculate the mutual information;

[0024] (2.1.3) Normalize the mutual information to obtain the normalized mutual information value;

[0025] (2.1.4) Traverse all network partitioning methods and calculate the normalized mutual information value under each partitioning method to calculate the maximum information coefficient MIC.

[0026] Preferably, step (3) specifically includes:

[0027] Define the sample dataset E = {X1, X2, ..., X...} n ,Y}, where X i Let Y be the dimension of the input variables after feature filtering, Y be the target variable (temperature of the sensitive plate in the distillation column), and the delay time range be set to τ = {0, 1, 2, ..., τ}. max}, where τ max The maximum delay time; for each feature variable X in the sample dataset E. i (i = 1, 2, ..., n), calculate the mutual information coefficient I(X) between it and the target variable Y at each delay time τ. i (t-τ); Y(t)), the delay time corresponding to the maximum information coefficient MIC between the auxiliary variable and the target variable is selected as the optimal delay time.

[0028] Preferably, step (4) specifically includes:

[0029] Calculate the number of time-delayed samples for each variable, and divide the optimal delay time of each auxiliary variable by the sampling time to obtain the number of time-delayed samples. Then, convert the original time series data into a data matrix, where the number of rows is the original number of rows minus the maximum number of time-delayed samples, and the number of columns is the number of auxiliary variables plus the number of target variables. Shift the original data according to the calculated number of time-delayed samples and fill it into the new matrix. Start filling the new matrix with the data of the target variables from the row with the maximum number of time-delayed samples to obtain a new dataset.

[0030] Preferably, step (5) specifically includes the following steps:

[0031] (5.1) Define the input layer, LSTM layer, activation layer, fully connected layer, and regression layer of the LSTM model;

[0032] (5.2) Select the optimizer, set the maximum number of iterations, set the initial learning rate, and configure the learning rate descent strategy to perform model parameter configuration processing;

[0033] (5.3) The acquired data is randomly divided into training set and test set, and the data is divided according to a preset ratio.

[0034] (5.4) Use the configured training parameters and preprocessed training data to train the model, and minimize the loss function of the model through iterative optimization;

[0035] (5.5) Calculate the evaluation index on the test set to evaluate the prediction accuracy and performance of the model and complete the model validation.

[0036] Preferably, step (6) specifically includes the following steps:

[0037] (6.1) Use the LSTM model described above to output the predicted temperature of the sensitive plate of the distillation column at future time.

[0038] (6.2) Design the objective function J(k) of the model predictive control MPC described above:

[0039]

[0040] Among them, y k Let y represent the predicted temperature at time k. ref Indicates the set temperature, △u k N represents the change in control input at time k, λ represents the weighting coefficient of the change in control input, and N represents the change in control input at time k. p N represents the length of the prediction time domain. c Indicates the length of the control time domain.

[0041] (6.3) Using the LSTM temperature prediction model, calculate the temperature prediction value of the MPC controller in the prediction time domain, and substitute it into the objective function for optimization to obtain the optimal control input sequence.

[0042] Preferably, step (7) specifically includes:

[0043] The nonlinear constrained optimization problem is solved using the particle swarm optimization algorithm, and the optimal control sequence is obtained through real-time rolling optimization. The first control input u in the optimal control input sequence is... t It is applied to the system, updates the system state, and enters the next control cycle, thereby completing the predictive control process.

[0044] The main feature of this LSTM-based model predictive control device for the temperature of the sensitive plate in a distillation column is that the device comprises:

[0045] A processor is configured to execute computer-executable instructions;

[0046] The memory stores one or more computer-executable instructions, which, when executed by the processor, implement the steps of the model predictive control method for the sensitive plate temperature of the distillation column based on LSTM described above.

[0047] The main feature of this LSTM-based model predictive control processor for the sensitive plate temperature of a distillation column is that the processor is configured to execute computer-executable instructions, which, when executed by the processor, implement the steps of the aforementioned LSTM-based model predictive control method for the sensitive plate temperature of a distillation column.

[0048] The computer-readable storage medium is characterized in that it stores a computer program that can be executed by a processor to implement the steps of the model predictive control method for the temperature of the sensitive plate of a distillation column based on LSTM described above.

[0049] The present invention employs the LSTM-based model predictive control method, device, processor, and computer-readable storage medium for the temperature of the sensitive plate in a distillation column. By constructing a model predictive controller for the temperature of the sensitive plate in a distillation column, the problems of poor overall performance and poor anti-interference ability of traditional PID controllers are solved. Simulation analysis using Simulink and Aspen Dynamics shows that the designed control method has very good dynamic characteristics, thus verifying the effectiveness and practicality of the control method. Attached Figure Description

[0050] Figure 1 This is a flowchart of the model predictive control method for the temperature of the sensitive plate in a distillation column based on LSTM, as described in this invention.

[0051] Figure 2 This is a distribution diagram of the MIC values ​​of the main process variables under different delay times in the model predictive control method for the temperature of the sensitive plate of the distillation column based on LSTM of the present invention.

[0052] Figure 3 This is a test set prediction result diagram of the LSTM prediction model trained by the model prediction control method for sensitive plate temperature of distillation column based on LSTM of the present invention.

[0053] Figure 4 This is a schematic diagram of the control strategy of the model predictive control method in the LSTM-based model predictive control method for sensitive plate temperature of a distillation column according to the present invention.

[0054] Figure 5 The figure shows the simulation results of the model predictive control method for the temperature of the sensitive plate of the distillation column based on LSTM according to the present invention. Detailed Implementation

[0055] To more clearly describe the technical content of the present invention, the following description is provided in conjunction with specific embodiments.

[0056] Before describing the embodiments of the present invention in detail, it should be noted that, in the following, the terms “comprising,” “including,” or any other variations are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed or inherent to such process, method, article, or apparatus.

[0057] Please see Figure 1 As shown, this LSTM-based model predictive control method for the temperature of the sensitive plate in a distillation column includes the following steps:

[0058] Step 1: Sample the data of the actual production process according to the sampling period to obtain the production history data of the distillation column in the actual factory production.

[0059] Step 2: Calculate the maximum information coefficient (MIC) between multiple variables and the sensitive plate temperature at different delay times. Use the MIC value of the data as a measure of the correlation between the feature variables and the sensitive plate temperature of the distillation column. Perform variable screening, eliminating feature variables with a maximum information coefficient lower than a set threshold, and determine the auxiliary variables used for sensitive plate temperature prediction. Take the delay time corresponding to the auxiliary variable reaching its maximum MIC as the optimal delay time between the auxiliary variable and the sensitive plate temperature, and then reconstruct the time delay of the original data based on the optimal delay time.

[0060] Step 3: Use the reconstructed dataset after time delay as training samples to build and train the LSTM model.

[0061] Step 4: Use the LSTM model trained in Step 3 as the prediction model in the model predictive controller. Consider the mechanical limitations of the distillation column and the upper and lower limits of variables in the process. Establish the objective optimization function based on the predicted output of the LSTM model and the reference value in the prediction time domain. Use the selected auxiliary variables as the input of the LSTM prediction model to predict the future state of the temperature of the sensitive plate of the distillation column. Use the particle swarm optimization algorithm (PSO) to solve the nonlinear constraint optimization problem. Then obtain the optimal control quantity through real-time rolling optimization.

[0062] In a preferred embodiment of the present invention, step 1 specifically includes the following steps:

[0063] (1.1) The production data of the distillation process in the factory is obtained by sensors and data acquisition system, and the raw data arranged in time axis is obtained, including multiple variables such as distillation column inlet flow rate, outlet flow rate, reboiler heat carrier flow rate, reflux flow rate and sensitive plate temperature.

[0064] In a preferred embodiment of the present invention, step 2 specifically includes the following steps:

[0065] (2.1) Calculate the maximum information coefficient (MIC) between the feature variables and the target variable in the production data collected in step 1 under different delay times. The maximum information coefficient (MIC) is calculated in the following way:

[0066] (2.1.1) Data discretization:

[0067] Suppose we have two continuous variables, X and Y, each with n observations. We divide the ranges of X and Y into several intervals of equal width or frequency. Generally, the number of intervals B should be less than n, and B is n. 0.6 For each variable, different partitioning methods are chosen to form different two-dimensional grids.

[0068] (2.1.2) Calculate mutual information:

[0069] For each partitioning method, data points are assigned to different cells of a two-dimensional grid. The number of data points in each cell is then calculated, forming a two-dimensional frequency table. The mutual information I(X;Y) is calculated using this frequency table.

[0070]

[0071] Where p(x,y) is the joint probability distribution, and p(x) and p(y) are the marginal probability distributions.

[0072] (2.1.3) Normalized mutual information:

[0073] To ensure comparability of mutual information values ​​under different grid partitioning methods, the mutual information is normalized. The normalization method typically involves dividing the mutual information by the logarithm of the number of grid partitions, i.e.:

[0074]

[0075] Where |X| and |Y| are the ranges of values ​​for X and Y, respectively.

[0076] (2.1.4) Calculate the maximum information coefficient (MIC):

[0077] Iterate through all possible mesh partitioning methods and calculate the normalized mutual information value for each partitioning method.

[0078] The maximum value of the normalized mutual information is selected as the maximum information coefficient (MIC):

[0079]

[0080] (2.2) Randomly generate a variable whose value is random, calculate the MIC between the random variable and the target variable, use the MIC of the random variable as a threshold, remove feature variables whose maximum information coefficient is lower than the threshold, perform preliminary variable screening on multiple feature variables in the production data, and obtain auxiliary variables for predicting temperature.

[0081] (2.3) Based on the maximum information coefficient (MIC) between the auxiliary variable and the target variable temperature, determine the auxiliary variable x. i The time delay parameter D between the target variable y(k) and the target variable y(k) i , where i = 1, 2, ..., m represents the number of auxiliary variables.

[0082] (2.4) Based on the optimal delay time between the auxiliary variable and the target variable, the collected process production data is reconstructed by time delay to obtain the dataset after filtering and time delay reconstruction in step (2.2). The specific steps for time delay reconstruction are as follows:

[0083] For each time step k, from max(D) i From x1 to x2, where T is the total number of time steps in the time series data, construct a single row containing the values ​​of all auxiliary variables after all time delays. For example, suppose there are m auxiliary variables x1, x2, ..., xn. m and the corresponding time delay parameters D1, D2, ... D m Given the total number of time steps T, the number of rows in the delay matrix will be T-max(D). i )+1, where the number of columns is the number of auxiliary variables m. The time delay matrix has the following form:

[0084]

[0085] In this matrix, each row corresponds to a time step, and each column corresponds to an auxiliary variable after a time delay.

[0086] In a preferred embodiment of the present invention, the steps for establishing the LSTM neural network model in step 3 are as follows:

[0087] (3.1) Model structure definition: Define the input layer, LSTM layer, activation layer, fully connected layer, regression layer, etc.

[0088] (3.2) Configure model parameters: Select optimizer, set maximum number of iterations, set initial learning rate, configure learning rate descent strategy, etc.

[0089] (3.3) Data partitioning: The data is randomly divided into training set and test set, with 80% training set and 20% test set.

[0090] (3.4) Model training: The model is trained using the configured training parameters and preprocessed training data. The model is optimized iteratively to minimize the loss function.

[0091] (3.5) Model validation: Calculate evaluation metrics on the test set, such as mean absolute error (MAE) and mean evaluation error (MBE), to assess the model’s prediction accuracy and performance.

[0092] In a preferred embodiment of the present invention, the steps for designing the MPC controller in step 4 are as follows:

[0093] The upper and lower temperature limits of the distillation column are set. Temperature data predicted by the LSTM temperature prediction model trained in step 3 after a future time step is input into the cost function. Within each sliding time window, rolling optimization is performed within the finite time domain [t, t+k]. The process of solving the optimal control system problem can be transformed into an optimization process of the cost function. The PSO algorithm is used to solve this optimization problem to find the optimal control sequence. These control variables are treated as particles in a multidimensional search space. By iteratively updating the particle's movement speed and direction, the optimal solution is obtained. The first control input u in the optimal control sequence is... t It is applied to the system, updates the system status, and initiates the next control cycle.

[0094] In a specific embodiment of the present invention, the specific implementation process of the method is as follows:

[0095] 1. Data acquisition and processing

[0096] In this embodiment, the raw data is the production process data of a pressure swing distillation plant. By analyzing the reaction mechanism of the distillation process, eight variables affecting the temperature of the sensitive plate in the distillation column were initially selected as characteristic variables, with the temperature of the sensitive plate as the target variable. The main process variables are shown in Table 1. 5000 operational data points were collected from the plant during operation, with a sampling period of 5 seconds.

[0097] Table 1. Main process variables of the distillation column

[0098]

[0099] The specific steps for calculating the maximum information coefficient (MIC) between the feature variables and the target variable in the collected production data at different delay times are as follows:

[0100] (1) Collect production data {X 5000×10 ,Y 5000×1 Each auxiliary variable X i The range of values ​​for Y is divided into several intervals of equal width or equal frequency, and the number of intervals B is n. 0.6 , where n is the total amount of data collected, 5000.

[0101] (2) Calculate mutual information

[0102] For each partitioning method, data points are assigned to different cells of a two-dimensional grid. The number of data points in each cell is then calculated to form a two-dimensional frequency table. Mutual information I(X;Y) is calculated using the following formula (1):

[0103]

[0104] Where p(x,y) is the joint probability distribution, and p(x) and p(y) are the marginal probability distributions.

[0105] (3) Normalized mutual information:

[0106] In order to make the mutual information values ​​under different partitioning methods comparable, the mutual information is normalized by the following formula (2).

[0107]

[0108] Where |X| and |Y| are the ranges of values ​​for X and Y, respectively.

[0109] (4) Calculate the maximum information coefficient (MIC):

[0110] Iterate through all possible mesh partitioning methods and calculate the normalized mutual information value for each partitioning method.

[0111] The maximum value of the normalized mutual information is selected as the maximum information coefficient (MIC) by the following formula (3):

[0112]

[0113] (5) Selection of auxiliary variables:

[0114] A random variable is generated, and its value is random. The minimum acceptable value (MIC) between this random variable and the target variable is calculated. The result is as follows: Figure 2 As shown, the result is 0.334. Setting the MIC threshold to 0.334, and removing feature variables below the threshold, variable filtering was performed on the feature variables in the production data to obtain auxiliary variables for temperature prediction. The MIC of each auxiliary variable after filtering and the optimal delay time between each auxiliary variable and the target variable are shown in Table 2.

[0115] Table 2

[0116]

[0117] (6) Delay reconstruction:

[0118] The delay time corresponding to the maximum MIC of the auxiliary variable is taken as the optimal delay time between the auxiliary variable and the temperature of the sensitive plate. The number of delay samples D for each variable is calculated. i The number of time-delayed samples is obtained by dividing the optimal delay time of each auxiliary variable by the sampling time. To convert the original time series data into a data matrix, the number of rows is the original number of rows minus the maximum number of time-delayed samples (max(D)). i The number of columns represents the number of auxiliary variables. The original data is shifted based on the calculated number of time-delay samples and then filled into a new matrix. The target variable data is then filled into the new matrix starting from the row with the largest number of time-delay samples. The time-delay matrix has the following form:

[0119]

[0120] 2. Construct and train the LSTM prediction model

[0121] (1) Constructing and training an LSTM neural network model

[0122] The LSTM neural network model consists of an input layer, an LSTM layer, a ReLU activation layer, a fully connected layer, and an output layer. The Adam optimizer was chosen, and after multiple experiments and parameter adjustments, the learning rate for the Adam algorithm was set to 0.01. 80% of the preprocessed data was used as the training set, and 20% as the test set. After numerous experiments, it was found that the model performed best with 1000 training iterations, so the maximum number of training iterations was set to 1000. The prediction results of the trained model on the test set are shown in the figure below. Figure 3 .

[0123] 3. Design an MPC temperature controller

[0124] The task of the distillation column sensitive plate temperature control model is to determine N c Given an input sequence, the predicted value [y] for the next sampling period is... p (k+1),y p (k+2),…,y p (k+N p The goal is to continuously approach the desired temperature of the distillation column. Solving the optimal control sequence problem can be transformed into the following equation (8): the objective function is an optimization process with equation (9) as the constraint. Here, the PSO algorithm is used to obtain the optimal control sequence u = [u(k), u(k+1), ..., u(k+N)]. c -1)], taking the first one from the optimal control sequence as the control input to the system. In this example, the control time domain N p The value is 5, and the prediction time domain N is... c It is 15.

[0125]

[0126] 4. Simulation Experiment

[0127] The final simulation experiment was based on the actual factory's production process parameters and requirements. AspenDynamic was used to build a dynamic simulation of the actual factory to simulate the production process. The simulation verification involved the following steps:

[0128] (1) Aspen Dynamic was used to obtain historical data of multiple characteristic variables in the distillation column production process by adding disturbances and changing set values ​​during the operation. The sampling time was 1 second, and 5000 time series data were collected.

[0129] (2) In MATLAB, the obtained time series data are processed in sequence according to the above steps to calculate MIC, select auxiliary variables and reconstruct data delay, build and train LSTM prediction model.

[0130] (3) Based on the trained LSTM prediction model, an LSTM model prediction controller is designed, and the control loop is built in Simulink. The control strategy is as follows: Figure 4 As shown, Aspen Dynamic simulation is used as the actual production site for co-simulation, and compared with conventional PID cascade control. During the simulation, the PID controller is adjusted to the optimal parameters to obtain the best control effect of this comparison method. The final control effect is compared to, for example... Figure 5 As shown.

[0131] This LSTM-based model predictive control device for the temperature of the sensitive plate in a distillation column includes:

[0132] A processor is configured to execute computer-executable instructions;

[0133] The memory stores one or more computer-executable instructions, which, when executed by the processor, implement the steps of the model predictive control method for the sensitive plate temperature of the distillation column based on LSTM described above.

[0134] The LSTM-based model predictive control processor for the temperature of the sensitive plate in a distillation column is configured to execute computer-executable instructions, which, when executed by the processor, implement the steps of the LSTM-based model predictive control method for the temperature of the sensitive plate in a distillation column.

[0135] The computer-readable storage medium contains a computer program that can be executed by a processor to implement the steps of the model predictive control method for the temperature of the sensitive plate of a distillation column based on LSTM described above.

[0136] Any process or method description in the flowchart or otherwise herein can be understood as representing a module, segment, or portion of code comprising one or more executable instructions for implementing a particular logical function or process, and the scope of the preferred embodiments of the invention includes additional implementations in which functions may be performed not in the order shown or discussed, including substantially simultaneously or in reverse order depending on the functions involved, as will be understood by those skilled in the art to which embodiments of the invention pertain.

[0137] It should be understood that various parts of the present invention can be implemented using hardware, software, firmware, or a combination thereof. In the above embodiments, multiple steps or methods can be implemented using software or firmware stored in memory and executed by a suitable instruction execution device.

[0138] Those skilled in the art will understand that all or part of the steps of the methods in the above embodiments can be implemented by a program instructing related hardware. The program can be stored in a computer-readable storage medium, and when executed, the program includes one or a combination of the steps of the method embodiments.

[0139] The storage media mentioned above can be read-only memory, disk, or optical disk, etc.

[0140] In the description of this specification, references to terms such as "an embodiment," "some embodiments," "example," "specific example," or "embodiment," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of the invention. In this specification, illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples.

[0141] Although embodiments of the present invention have been shown and described above, it is understood that the above embodiments are exemplary and should not be construed as limiting the present invention. Those skilled in the art can make changes, modifications, substitutions and variations to the above embodiments within the scope of the present invention.

[0142] The present invention employs the LSTM-based model predictive control method, device, processor, and computer-readable storage medium for the temperature of the sensitive plate in a distillation column. By constructing a model predictive controller for the temperature of the sensitive plate in a distillation column, the problems of poor overall performance and poor anti-interference ability of traditional PID controllers are solved. Simulation analysis using Simulink and Aspen Dynamics shows that the designed control method has very good dynamic characteristics, thus verifying the effectiveness and practicality of the control method.

[0143] In this specification, the invention has been described with reference to specific embodiments thereof. However, it will be apparent that various modifications and variations can be made without departing from the spirit and scope of the invention. Therefore, the specification and drawings should be considered illustrative rather than restrictive.

Claims

1. A model predictive control method for the temperature of the sensitive plate in a distillation column based on LSTM, characterized in that, The method includes the following steps: (1) Collect historical data of multiple variables during the production process of the distillation column to obtain the original process data arranged according to the time axis; (2) Calculate the maximum information coefficient MIC of multiple variables under different delay times, use the MIC value as an evaluation index of the correlation between auxiliary variables and the temperature of the sensitive plate of the distillation column, and determine the auxiliary variables required to predict the temperature of the sensitive plate through variable screening. (3) The delay time corresponding to the maximum MIC value of the auxiliary variable is taken as the optimal delay time between the auxiliary variable and the temperature of the sensitive plate; (4) Based on the optimal delay time, the original data is reconstructed to obtain a new dataset; (5) Use the dataset obtained after time delay reconstruction as the model training sample to build and train the LSTM model; (6) The trained LSTM model is used as the real-time prediction model of the model predictive control (MPC) to design a sensitive plate temperature controller. (7) Establish the objective optimization function based on the predicted output of the LSTM model and the reference value in the prediction time domain, and obtain the optimal control quantity predicted by the model by solving the nonlinear constraint optimization problem.

2. The model predictive control method for sensitive plate temperature of a distillation column based on LSTM according to claim 1, characterized in that, The specific steps (1) are as follows: Historical data on the production process of the distillation column were collected from the factory, including multiple variables such as the inlet flow rate, outlet flow rate, reboiler heat carrier flow rate, column top reflux flow rate, and distillation column sensitive plate temperature. The data were sampled according to the sampling period to obtain the raw process data arranged on the time axis.

3. The model predictive control method for sensitive plate temperature of a distillation column based on LSTM according to claim 2, characterized in that, Step (2) specifically includes the following steps: (2.1) Calculate the maximum information coefficient (MIC) between the feature variables and the target variable in the collected historical production process data under different delay times; (2.2) Take the MIC between each feature variable and the target variable as the correlation index, remove feature variables with MIC below the set threshold, perform preliminary variable screening on multiple feature variables in the production data, and obtain auxiliary variables for predicting the temperature of the sensitive plate of the distillation column.

4. The model predictive control method for sensitive plate temperature of a distillation column based on LSTM according to claim 3, characterized in that, Step (2.1) specifically includes the following steps: Suppose we have two variables, X and Y, containing n measurements. (2.1.1) Divide variables X and Y into several intervals and form a two-dimensional grid to achieve data discretization; (2.1.2) For each grid, count the number of data points in each cell to form a two-dimensional frequency table and calculate the mutual information; (2.1.3) Normalize the mutual information to obtain the normalized mutual information value; (2.1.4) Traverse all network partitioning methods and calculate the normalized mutual information value under each partitioning method to calculate the maximum information coefficient MIC.

5. The model predictive control method for sensitive plate temperature of a distillation column based on LSTM according to claim 4, characterized in that, Step (3) specifically refers to: Define the sample dataset E = {X1, X2, ..., X...} n ,Y}, where n is the dimension of the input variable after feature filtering, Y is the target variable, the temperature of the sensitive plate of the distillation column, and the delay time range is set to τ={0,1,2,...,τ}. max }, where τ max The maximum delay time; for each feature variable X in dataset E. i (i = 1, 2, ..., n), calculate the mutual information coefficient I(X) between it and the target variable Y at each delay time τ. i (t-τ); Y(t)), the delay time corresponding to the maximum information coefficient MIC between the auxiliary variable and the target variable is selected as the optimal delay time.

6. The model predictive control method for sensitive plate temperature of a distillation column based on LSTM according to claim 5, characterized in that, Step (4) specifically refers to: Calculate the number of time-delayed samples for each variable, and divide the optimal delay time of each auxiliary variable by the sampling time to obtain the number of time-delayed samples. Then, convert the original time series data into a data matrix, where the number of rows is the original number of rows minus the maximum number of time-delayed samples, and the number of columns is the number of auxiliary variables plus the number of target variables. Shift the original data according to the calculated number of time-delayed samples and fill it into the new matrix. Start filling the new matrix with the data of the target variables from the row with the maximum number of time-delayed samples to obtain a new dataset.

7. The model predictive control method for sensitive plate temperature of a distillation column based on LSTM according to claim 6, characterized in that, Step (5) specifically includes the following steps: (5.1) Define the input layer, LSTM layer, activation layer, fully connected layer, and regression layer of the LSTM model; (5.2) Select the optimizer, set the maximum number of iterations, set the initial learning rate, and configure the learning rate descent strategy to perform model parameter configuration processing; (5.3) The acquired data is randomly divided into training set and test set, and the data is divided according to a preset ratio; (5.4) Use the configured training parameters and preprocessed training data to train the model, and minimize the loss function of the model through iterative optimization; (5.5) Calculate the evaluation index on the test set to evaluate the prediction accuracy and performance of the model and complete the model validation.

8. The model predictive control method for sensitive plate temperature of a distillation column based on LSTM according to claim 7, characterized in that, Step (6) specifically includes the following steps: (6.1) Use the LSTM model described above to output the predicted temperature of the sensitive plate of the distillation column at future time. (6.2) Design the objective function J(k) of the model predictive control MPC described above: Among them, y k Let y represent the predicted temperature at time k. ref Indicates the set temperature, △u k N represents the change in control input at time k, λ represents the weighting coefficient of the change in control input, and N represents the change in control input at time k. p N represents the length of the prediction time domain. c Indicates the length of the control time domain; (6.3) Using the LSTM temperature prediction model, calculate the temperature prediction value of the MPC controller in the prediction time domain, and substitute it into the objective function for optimization to obtain the optimal control input sequence.

9. The model predictive control method for sensitive plate temperature of a distillation column based on LSTM according to claim 8, characterized in that, The specific steps (7) are as follows: The nonlinear constrained optimization problem is solved using the particle swarm optimization algorithm, and the optimal control sequence is obtained through real-time rolling optimization. The first control input u in the optimal control input sequence is... t It is applied to the system, updates the system state, and enters the next control cycle, thereby completing the predictive control process.

10. A model predictive control device for the temperature of a sensitive plate in a distillation column based on LSTM, characterized in that, The device includes: A processor is configured to execute computer-executable instructions; The memory stores one or more computer-executable instructions, which, when executed by the processor, implement the steps of the model predictive control method for sensitive plate temperature of a distillation column based on LSTM, as described in any one of claims 1 to 9.

11. A model predictive control processor for the temperature of a sensitive plate in a distillation column based on LSTM, characterized in that, The processor is configured to execute computer-executable instructions, which, when executed by the processor, implement the steps of the model predictive control method for sensitive plate temperature of a distillation column based on LSTM as described in any one of claims 1 to 9.

12. A computer-readable storage medium, characterized in that, It stores a computer program that can be executed by a processor to implement the steps of the model predictive control method for the temperature of the sensitive plate of a distillation column based on LSTM, as described in any one of claims 1 to 9.

Images