Establishment method of intelligent control model of vacuum concentration system

Abstract

The invention relates to a method for establishing an intelligent control model of a vacuum concentration system, and belongs to the technical field of intelligence. Comprising the following steps: data acquisition: acquiring equipment temperature T, jacket pressure P, vacuum degree Ps and system stability S in a vacuum concentration system; bP-ANN neural network modeling: an artificial neural network based on a BP algorithm is selected to establish a prediction model, and the model takes the equipment temperature T and the jacket pressure P as input and takes the predicted vacuum degree Ps and the system stability S as output. The model provided by the invention has high prediction precision, and lays a foundation for the design of a control strategy; through comparison experiment results of neural network model predictive control and PID control, the predictive control is obviously superior to the PID control in the aspect of stability, and the superiority of a BP-ANN model predictive control scheme is further verified.

Description

Technical Field

[0001] The invention relates to a method for establishing an intelligent control model of a vacuum concentration system, and belongs to the field of intelligent technology. Background Art

[0002] Automatic control of industrial production processes plays an important role in many industrial fields. The quality of control directly affects the quality, stability and uniformity of products, as well as the safety, stability and labor intensity of workers. Although some Chinese herbal medicine extractions currently use process control systems, many problems have not been effectively solved. For simple process systems, satisfactory control effects can be obtained by using traditional classical or modern control theory methods. However, for complex process systems such as Chinese herbal medicine extraction processes, due to their uncertain characteristics such as nonlinearity, time-varying, hysteresis, and multivariable coupling, it is difficult to obtain good control effects using traditional control methods. Therefore, intelligent control methods have become a research hotspot.

[0003] Intelligent control methods include expert system control, fuzzy control, neural network control and human-like intelligent control, etc. These methods have their own characteristics and have been successfully applied in different fields. With the improvement of the automation level of the industry, it is very necessary to establish an intelligent control model of vacuum concentration system according to the process characteristics of the current Chinese medicine extraction industry.

[0004] Therefore, it is urgent to propose a method for establishing an intelligent control model of a vacuum concentration system to solve the above technical problems. Summary of the invention

[0005] In order to solve the above problems, a method for establishing an intelligent control model of a vacuum concentration system is provided. A brief overview of the present invention is given below to provide a basic understanding of certain aspects of the present invention. It should be understood that this overview is not an exhaustive overview of the present invention. It is not intended to determine the key or important parts of the present invention, nor is it intended to limit the scope of the present invention.

[0006] The technical solution of the present invention:

[0007] A method for establishing an intelligent control model of a vacuum concentration system comprises the following steps:

[0008] Step 1: Data Collection

[0009] In the vacuum concentration system, data including equipment temperature T, jacket pressure P, vacuum degree Ps, and system stability S were collected;

[0010] Step 2: BP-ANN neural network modeling

[0011] An artificial neural network based on BP algorithm was selected to establish a prediction model, which takes equipment temperature T and jacket pressure P as input and predicts vacuum degree Ps and system stability S as output.

[0012] Preferably: Step 1 comprises the following steps:

[0013] Step 1.1: Record the real-time changes of equipment temperature T and jacket pressure P through high-precision sensors;

[0014] Step 1.2: Use a vacuum pressure sensor to measure the vacuum degree Ps;

[0015] Step 1.3: Define and quantify stability S based on the system oscillation level and key performance indicators;

[0016] Step 1.4: The data collection frequency is set to once every 5 seconds, covering the dynamic characteristics of the system under different operating modes.

[0017] Preferably: Step 2 comprises the following steps:

[0018] Step 2.1: Data preprocessing;

[0019] Step 2.2: Network architecture design;

[0020] Step 2.3: Model training.

[0021] Preferably: in step 2.1, data standardization is performed to ensure that the input variables and output variables are within similar numerical ranges, thereby improving the convergence speed and prediction accuracy of the model;

[0022] Divide the data into training, validation, and test sets to ensure that the model has good generalization ability;

[0023] In step 2.2, the input layer contains M nodes;

[0024] The hidden layer is set to 1 layer, each layer contains 10 neurons, and the activation function uses the positive and negative symmetric Tanh function;

[0025] The output layer contains 2 nodes, corresponding to the predicted vacuum degree Ps and stability S, respectively, and the activation function is a non-negative Tanh function;

[0026] In step 2.3, the loss function uses mean square error (MSE);

[0027] The optimizer selects the Adam algorithm and the learning rate is set to 0.001;

[0028] During the training process, hyperparameters are adjusted dynamically, and the overfitting of the model is monitored in real time through the validation set.

[0029] Preferably, the method further comprises step 3: neural network model predictive control, comprising the following steps:

[0030] Step 3.1: Predict the vacuum degree Ps and stability S in real time through the BP-ANN model, compare them with the set target values, and calculate the feedforward deviation;

[0031] Step 3.2: Fusion of feedforward and feedback biases via λ∈[0,1];

[0032] Step 3.3: After the total error is input into the controller, the output is the opening of the regulating jacket regulating valve, the number of vacuum pump operations and the number of cooling water pump operations.

[0033] Preferably, step 3 also includes step 3.4: comparing the neural network model predictive control with the PID control, in order to evaluate the predictive control effect based on the BP-ANN model.

[0034] Preferably: in step 3.4, the conventional PID control is to adjust the jacket regulating valve opening (TCV), the number of vacuum pump operations (Np), and the number of cooling water pump operations (Nl) according to the set target vacuum degree Ps and stability S; the error is adjusted, and the parameters are optimized by manual adjustment;

[0035] By comparing the experimental results of neural network model predictive control and PID control, the predictive control is significantly better than PID control in terms of stability, which further verifies the superiority of the BP-ANN model predictive control scheme.

[0036] Preferred: Application to an artificial intelligence based vacuum concentration system.

[0037] The present invention has the following beneficial effects:

[0038] 1. The present invention clearly displays the system behavior, the changing trends of equipment temperature, jacket pressure, vacuum degree and stability; by comparing the predicted vacuum degree and stability, the model has a high prediction accuracy, which lays a foundation for the design of control strategy; by comparing the experimental results of neural network model predictive control and PID control, the predictive control is significantly better than PID control in terms of stability, which further verifies the superiority of the BP-ANN model predictive control scheme;

[0039] 2. When the air duct is opened, the air is heated by the steam in the gap, thereby eliminating foam and reducing the impact on evaporation efficiency, realizing the reuse of steam and improving energy utilization; the raw material tank of the present invention is used to buffer the cached drug solution, ensure the continuity of the process, improve efficiency, and is suitable for large-size meter-level standards; the height of the evaporation chamber and the gas-liquid separator of the present invention and the arrangement of the conduit not only recover the condensate containing the drug components to a great extent, but also effectively prevent it from entering the condenser, thereby improving the collection purity of the solvent and reducing the loss of the drug components; after the first-stage condenser and the second-stage condenser are condensed once and twice, the steam temperature is still around 80°C. Due to the different boiling points of the solvents, such as the low boiling point of alcohol, 2-7°C chilled water is introduced into the third-stage condenser for condensation to improve the recovery efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] Figure 1 It is a diagram of a method for establishing an intelligent control model of a vacuum concentration system;

[0041] Figure 2 This is the trend chart of the first 1000 sets of data;

[0042] Figure 3 This is a comparison chart of vacuum degree and stability predicted by the neural network;

[0043] Figure 4 It is a predictive control control flow chart;

[0044] Figure 5 It is the PID control flow chart;

[0045] Figure 6 This is a comparison chart of PID and model predictive control effects;

[0046] Figure 7 It is an intelligent control model;

[0047] Figure 8 A block diagram of an artificial intelligence-based vacuum concentration system;

[0048] Fig. 9 It is a schematic diagram of the structure of the heating chamber;

[0049] Fig.10 is a schematic diagram of the structure of the evaporation chamber;

[0050] Fig.11 It is a schematic diagram of the structure of the condenser.

[0051] In the figure: 1-heating chamber, 2-evaporation chamber, 3-condenser, 4-primary condenser, 5-secondary condenser, 6-tertiary condenser, 7-liquid collector, 8-water pump, 9-gas-liquid separator, 10-conduit, 11-elbow, 12-jacket, 13-air guide tube, 14-liquid guide tube, 15-density sensor, 16-temperature sensor, 17-chain electrode, 18-fixing ring, 19-pressure sensor, 20-condenser connecting pipe, 21-vacuum tube, 22-raw material tank, 41-primary cooling water inlet, 42-primary cooling water outlet, 51-secondary cooling water inlet, 52-secondary cooling water outlet, 61-tertiary cooling water inlet, 62-tertiary cooling water outlet, 71-upper Liquid level gauge, 72-lower liquid level gauge, 73-liquid collector sight glass, 74-liquid collector sight light mirror, 101-liquid outlet, 102-liquid inlet and outlet, 103-feeding port, 104-discharging port, 105-pipe drain valve, 106-heating chamber steam inlet, 107-steam pressure sensor, 108-heating chamber steam outlet, 121-jacket steam inlet, 122-jacket steam outlet, 201-steam outlet, 202-evaporation chamber outlet, 203-evaporation chamber inlet, 204-sight glass, 205-cleaning hole, 206-manhole, 207-light sight glass, 221-dosing tank pipeline, 222-cleaning water tank pipeline, 223-discharge pipeline, 224-transport pipeline. DETAILED DESCRIPTION

[0052] In order to make the purpose, technical scheme and advantages of the present invention clearer, the present invention is described below by the specific embodiments shown in the accompanying drawings. However, it should be understood that these descriptions are only exemplary and are not intended to limit the scope of the present invention. In addition, in the following description, the description of well-known structures and technologies is omitted to avoid unnecessary confusion of the concept of the present invention.

[0053] Specific implementation method 1: Combination Figure 1-11 The present embodiment is described. A method for establishing an intelligent control model of a vacuum concentration system in the present embodiment comprises the following steps:

[0054] Step 1: Data Collection

[0055] In the vacuum concentration system, operating data including equipment temperature T, jacket pressure P, vacuum degree Ps, and system stability S were collected, with a total of 60,000 sets of data obtained;

[0056] Step 1 includes the following specific collection process steps:

[0057] Step 1.1: Record the real-time changes of equipment temperature T and jacket pressure P through high-precision sensors;

[0058] Step 1.2: Use a vacuum pressure sensor to measure the vacuum degree Ps;

[0059] Step 1.3: Define and quantify stability S based on the system oscillation level and key performance indicators;

[0060] Step 1.4: The data collection frequency is set to once every 5 seconds, covering the dynamic characteristics of the system in different operating modes; in order to more clearly show the system behavior, such as Figure 2 As shown, the trend chart of the first 1000 sets of data is drawn to show the changing trends of equipment temperature, jacket pressure, vacuum degree and stability;

[0061] Step 2: BP-ANN neural network modeling

[0062] An artificial neural network (ANN) based on BP algorithm was selected to establish a prediction model, which takes equipment temperature T and jacket pressure P as input and predicts vacuum degree Ps and system stability S as output;

[0063] Step 2 includes the following modeling process steps:

[0064] Step 2.1: Data preprocessing;

[0065] In step 2.1, data standardization is performed to ensure that the input variables and output variables are within similar numerical ranges, thereby improving the convergence speed and prediction accuracy of the model;

[0066] The data was divided into training set (70%), validation set (15%) and test set (15%) to ensure that the model has good generalization ability;

[0067] Step 2.2: Network architecture design;

[0068] In step 2.2, the input layer contains M nodes (equipment temperature T and jacket pressure P);

[0069] The hidden layer is set to 1 layer, each layer contains 10 neurons, and the activation function uses the positive and negative symmetric Tanh function;

[0070] The output layer contains 2 nodes, corresponding to the predicted vacuum degree Ps and stability S, respectively, and the activation function is a non-negative Tanh function;

[0071] Step 2.3: Model training;

[0072] In step 2.3, the loss function uses mean square error (MSE);

[0073] The optimizer selects the Adam algorithm and the learning rate is set to 0.001;

[0074] Dynamically adjust hyperparameters during training and monitor the overfitting of the model in real time through the validation set;

[0075] like Figure 3 As shown in the figure, after the training is completed, a comparison chart of the predicted vacuum degree and stability is drawn; the results show that the model has a high prediction accuracy, which lays the foundation for the design of control strategy;

[0076] Step 3: Neural Network Model Predictive Control

[0077] The following steps are involved:

[0078] like Figure 4 As shown, step 3.1: predict the vacuum degree Ps and stability S in real time through the BP-ANN model, compare them with the set target value, and calculate the feedforward deviation;

[0079] Step 3.2: Fusion of feedforward and feedback biases via λ∈[0,1];

[0080] Step 3.3: After the total error is input into the controller, the output is the opening of the regulating jacket regulating valve, the number of vacuum pump operations and the number of cooling water pump operations;

[0081] Step 3.4: Comparison between neural network model predictive control and PID control. In order to evaluate the predictive control effect based on the BP-ANN model, a comparative experiment was designed to compare it with the traditional PID control method. 7. The method for establishing a vacuum concentration system intelligent control model according to claim 5 is characterized in that: in step 3.4, the traditional PID control is to adjust the jacket regulating valve opening (TCV), the number of vacuum pump operations (Np), and the number of cooling water pump operations (Nl) according to the set target vacuum degree Ps and stability S; using the traditional PID controller such as Figure 5 As shown, the error is adjusted and the parameters are optimized through manual adjustment;

[0082] The experimental results are compared by using neural network model predictive control and PID control, such as Figure 6 As shown in the figure, the predictive control is significantly better than the PID control in terms of stability, which further verifies the superiority of the BP-ANN model predictive control scheme;

[0083] Intelligence is stimulated and grown by objective information under the constraints of subject purpose and knowledge. It is a high-level product of the evolution of information ecology. By using the current rapidly developing software and hardware platform, the present invention establishes an intelligent control model of the vacuum concentration system. Figure 7 shown.

[0084] Knowledge is the basis for the conversion from information to strategy. Various types of knowledge in the system drive their respective functional modules, including common sense knowledge driving the perception information module (sensor data selection and generation) and the basic module (manual control); experience knowledge driving the intelligent module (prediction); normative knowledge driving the comprehensive decision-making module (automatic control). At the same time, various types of knowledge realize the internal ecological cycle of knowledge during the operation of the system.

[0085] Information conversion is the source of the law of intelligent creation; the information in the system is processed and converted in each module. The perception information module converts the sensor signal into grammatical information (system engineering quantity data), and the information conversion module converts the grammatical information into full information (data status, change and other characteristic quantities); the basic module, intelligent module and decision module convert grammatical information and full information into control strategies of different intelligence levels; at the same time, the information conversion process helps knowledge discovery and growth;

[0086] Software and hardware platform implementation: Software and hardware platforms are the implementation forms of system intelligent control. The system is divided into four levels of software and hardware platforms, including equipment environment layer (sensors and actuators), control layer (controllers and electrical circuits), intelligent layer (computers and computing software), database (data storage memory hard disk and software)

[0087] A method for establishing an intelligent control model of a vacuum concentration system is applied to a vacuum concentration system based on artificial intelligence, comprising a heating chamber 1, an evaporation chamber 2 and a condenser 3, wherein the heating chamber 1, the evaporation chamber 2 and the condenser 3 are connected in sequence, and further comprising a vapor-liquid separator 9, wherein the evaporation chamber 2 is connected to the condenser 3 through the vapor-liquid separator 9, and the vapor-liquid separator 9 is located above the evaporation chamber 2;

[0088] It also includes a 90° elbow 11, the upper end of the evaporation chamber 2 has a steam outlet 201, the condenser 3 has a condenser tube inlet, the condenser tube inlet and the vapor-liquid separator 9 are both located above the steam outlet 201, the inlet of the vapor-liquid separator 9 is connected to the steam outlet 201 through the elbow 11, and the outlet of the vapor-liquid separator 9 is connected to the condenser inlet of the condenser 3;

[0089] It also includes a conduit 10, which is an L-shaped pipe, including a horizontal pipe, an arc-shaped pipe and a vertical pipe. The horizontal pipe, the arc-shaped pipe and the vertical pipe are connected in sequence. The arc-shaped pipe is a 180° semicircular curved pipe, and the arc-shaped pipe is arranged in a downward concave direction. A small amount of liquid can be stored here to play a sealing role. The upper end of the vertical pipe of the conduit 10 is connected to the lower end discharge port of the vapor-liquid separator 9, and the horizontal pipe of the conduit 10 passes through the side wall of the evaporation chamber 2 and extends into the evaporation chamber 2. The horizontal pipe of the conduit 10 is located below the steam outlet 201, and a butterfly valve is arranged at the lowest point of the arc-shaped pipe for draining or replenishing liquid;

[0090] It also includes a jacket 12 and an air guide pipe 13. The outer side of the evaporation chamber 2 is wrapped with the jacket 12. The jacket 12 has a jacket steam inlet 121 on the upper side and a jacket steam outlet 122 on the lower side. There is a gap between the jacket 12 and the outer side of the evaporation chamber 2. The jacket steam inlet 121 and the jacket steam outlet 122 are both connected to the gap. One end of the air guide pipe 13 is connected to the air hole on the upper side of the evaporation chamber 2. The other end of the air guide pipe 13 passes through the gap and extends out. A pressure sensor 19 is provided on the air guide pipe 13 for detecting the vacuum degree of the evaporation chamber. When the valve on the air guide pipe is opened to adjust the pressure, the air is heated by the steam in the gap, thereby eliminating foam while reducing the impact on the evaporation efficiency. Through ingenious design, the energy utilization rate is improved under the premise of fully considering the overall layout during actual installation;

[0091] It also includes a temperature sensor 16, a chain electrode 17 and a fixing ring 18. The fixing ring 18 is arranged at the lower part of the evaporation chamber 2. The chain electrode 17 is inserted from the upper end through hole of the evaporation chamber 2 into the interior of the evaporation chamber 2 and connected with the fixing ring 18. The upper end of the evaporation chamber 2 has a cleaning hole 205 for installing a rotating nozzle. The upper end of the evaporation chamber 2 is installed with a light sight glass 207. The lower end of the evaporation chamber 2 has an evaporation chamber outlet 202. The side of the evaporation chamber 2 has an evaporation chamber inlet 203. The side of the evaporation chamber 2 has a sight glass 204 and a manhole 206. The sight glass 204 is located on the upper side of the manhole 206 for easy observation and maintenance. The lower side of the evaporation chamber 2 is provided with a temperature sensor 16. The evaporation chamber outlet 202 and the evaporation chamber inlet 203 are connected to the heating chamber 1.

[0092] The heating chamber 1 also includes a liquid guide tube 14 and a density sensor 15. The upper side of the heating chamber 1 has a liquid medicine outlet 101 connected to its shell side, the lower end of the heating chamber 1 has a liquid medicine inlet and outlet 102 connected to its shell side, the upper side of the heating chamber 1 has a heating chamber steam inlet 106 connected to its tube side, the lower side of the heating chamber 1 has a heating chamber steam outlet 108 connected to its tube side, the lower part of the heating chamber 1 has a tube side drain valve 105 connected to its tube side, the liquid medicine outlet 101 is connected to the evaporation chamber inlet 203, the evaporation chamber outlet 202 is connected to the liquid medicine inlet at the lower part of the heating chamber 1 through the liquid guide tube 14, the liquid medicine inlet and outlet 102 are connected to the dilute material feed port 103 and the concentrated liquid discharge port 104, the dilute material feed port 103 , a stop valve is respectively arranged between the concentrated liquid discharge port 104 and the liquid medicine inlet and outlet 102, the liquid guide pipe 14 is a 90° elbow, a density sensor 15 is arranged in the middle of the horizontal pipe of the liquid guide pipe 14 for detecting the concentration of the liquid medicine, the heating chamber steam inlet 106 is connected to the steam device for heating, a steam pressure sensor 107 is arranged at the steam outlet of the steam device, the outlet of the steam device is also connected to the jacket steam inlet 121, or the steam device, the heating chamber steam inlet 106, the pipe process, the heating chamber steam outlet 108, and the jacket steam inlet 121 are connected in sequence, so as to realize the reuse of steam, improve the utilization efficiency, and save energy, and the outer sides of the heating chamber 1, the liquid guide pipe 14, and the elbow 11 are all wrapped with a thermal insulation layer;

[0093] The condenser 3 includes a primary condenser 4, a secondary condenser 5, a tertiary condenser 6, a liquid collector 7 and a water pump 8. The outlet of the vapor-liquid separator 9 is connected to the shell side inlet of the upper part of the primary condenser 4. The shell side outlet of the lower part of the primary condenser 4 is connected to the shell side inlet of the upper part of the secondary condenser 5 through a condenser connecting pipe 20. The shell side outlet of the lower part of the secondary condenser 5 is connected to the shell side inlet of the tertiary condenser 6. The shell side outlet of the tertiary condenser 6 is connected to the shell side inlet of the lower part of the secondary condenser 5. The tube side of the primary condenser 4 has a primary cooling water inlet 41 and a primary cooling water outlet 42 for connecting a pump and a water tank. The cooling water tank, the water pump of the primary condenser, the primary cooling water inlet 41, the tube side of the primary condenser, and the primary cooling water outlet 42 are connected in a circular manner. The tube side of the secondary condenser 5 has a secondary cooling water inlet 51 and a secondary cooling water outlet 52 for connecting a pump and a water tank. The cooling water tank, the water pump of the secondary condenser, the secondary cooling water inlet 51, and the tube side of the secondary condenser , the secondary cooling water outlet 52 are connected in a circular manner, the cooling water tank is filled with normal temperature water, the freezing water tank is filled with 2-7°C water, the pipe side of the tertiary condenser 6 has a tertiary cooling water inlet 61 and a tertiary cooling water outlet 62 for connecting the pump and the water tank, the freezing water tank, the water pump of the tertiary condenser, the tertiary cooling water inlet 61, the pipe side of the tertiary condenser, and the tertiary cooling water outlet 62 are connected in a circular manner, and the condensed water outlets at the bottom of the primary condenser 4 and the secondary condenser 5 are connected to the liquid collector 7 below them. The front side of the liquid collector 7 is provided with a liquid collector sight glass 73, the rear side of the liquid collector 7 is provided with a liquid collector sight glass 74, the upper and lower sides of the liquid collector 7 are provided with an upper liquid level gauge 71 and a lower liquid level gauge 72, the lower end of the liquid collector 7 is provided with a drainage pipe, the drainage pipe is connected to the inlet of the water pump 8, the outlet of the water pump 8 is provided with a one-way valve 9, the third-stage condenser 6 adopts a threaded tube spiral winding cooler, the water pump 8 is a condensate pump, and the side of the second-stage condenser 5 is provided with a vacuum tube 21, and the vacuum tube 21 is connected to the vacuum pump;

[0094] It also includes a raw material tank 22, the upper opening of which is connected to a dosing tank pipeline 221 and a cleaning water tank pipeline 222 through valves, and the lower opening of the raw material tank 22 is connected to a discharge pipeline 223 and a delivery pipeline 224 through valves, and the delivery pipeline 224 is connected to the feed port 103. Valves are required to be set at the inlet and outlet of the present invention, and the raw material tank 22 is used for buffering and caching to ensure the continuity of the process and improve efficiency, and is suitable for large-size meter-level standards.

[0095] It should be noted that in the above embodiments, as long as the technical solutions are not contradictory, they can be arranged and combined, and those skilled in the art can exhaust all possibilities based on the mathematical knowledge of arrangement and combination. Therefore, the present invention will no longer describe the technical solutions after arrangement and combination one by one, but it should be understood that the technical solutions after arrangement and combination have been disclosed by the present invention.

[0096] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. For those skilled in the art, the present invention may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A method for establishing an intelligent control model of a vacuum concentration system, characterized in that: The following steps are involved: Step 1: Data Collection In the vacuum concentration system, data including equipment temperature T, jacket pressure P, vacuum degree Ps, and system stability S were collected; Step 2: BP-ANN neural network modeling An artificial neural network based on BP algorithm was selected to establish a prediction model, which takes equipment temperature T and jacket pressure P as input and predicts vacuum degree Ps and system stability S as output.

2. The method for establishing an intelligent control model of a vacuum concentration system according to claim 1, characterized in that: Step 1 includes the following steps: Step 1.1: Record the real-time changes of equipment temperature T and jacket pressure P through high-precision sensors; Step 1.2: Use a vacuum pressure sensor to measure the vacuum degree Ps; Step 1.3: Define and quantify stability S based on the system oscillation level and key performance indicators; Step 1.4: The data collection frequency is set to once every 5 seconds, covering the dynamic characteristics of the system under different operating modes.

3. The method for establishing a vacuum concentration system intelligent control model according to claim 2, characterized in that: Step 2 includes the following steps: Step 2.1: Data preprocessing; Step 2.2: Network architecture design; Step 2.3: Model training.

4. The method for establishing an intelligent control model of a vacuum concentration system according to claim 3, characterized in that: In step 2.1, data standardization is performed to ensure that the input variables and output variables are within similar numerical ranges, thereby improving the convergence speed and prediction accuracy of the model; Divide the data into training, validation, and test sets to ensure that the model has good generalization ability; In step 2.2, the input layer contains M nodes; The hidden layer is set to 1 layer, each layer contains 10 neurons, and the activation function uses the positive and negative symmetric Tanh function; The output layer contains 2 nodes, corresponding to the predicted vacuum degree Ps and stability S, respectively, and the activation function is a non-negative Tanh function; In step 2.3, the loss function uses mean square error; The optimizer selects the Adam algorithm and the learning rate is set to 0.001; During the training process, hyperparameters are adjusted dynamically, and the overfitting of the model is monitored in real time through the validation set.

5. The method for establishing an intelligent control model of a vacuum concentration system according to claim 4, characterized in that: Also included is step 3: neural network model predictive control, including the following steps: Step 3.1: Predict the vacuum degree Ps and stability S in real time through the BP-ANN model, compare them with the set target values, and calculate the feedforward deviation; Step 3.2: Fusion of feedforward and feedback biases via λ∈[0,1]; Step 3.3: After the total error is input into the controller, the output is the opening of the regulating jacket regulating valve, the number of vacuum pump operations and the number of cooling water pump operations.

6. The method for establishing an intelligent control model of a vacuum concentration system according to claim 5, characterized in that: Step 3 also includes step 3.4: comparison between neural network model predictive control and PID control, in order to evaluate the predictive control effect based on the BP-ANN model.

7. The method for establishing an intelligent control model of a vacuum concentration system according to claim 6, characterized in that: In step 3.4, the traditional PID control is to adjust the opening of the jacket regulating valve, the number of vacuum pump operations, and the number of cooling water pump operations according to the set target vacuum degree Ps and stability S; adjust the error, and optimize the parameters through manual adjustment; By comparing the experimental results of neural network model predictive control and PID control, the predictive control is significantly better than PID control in terms of stability, which verifies the superiority of the BP-ANN model predictive control scheme.

8. A method for establishing an intelligent control model for a vacuum concentration system according to any one of claims 1 to 7, characterized in that: Applied to an artificial intelligence-based vacuum concentration system.

Images