Leaf moistening and feeding control method and system, electronic equipment and computer medium

Abstract

The invention provides a leaf moistening and feeding control method and system, electronic equipment and a computer medium. The method comprises the following steps: collecting real-time data; based on the real-time data, a water adding amount reference value of the current time point is obtained; environment parameters of the current time point are obtained, the water adding amount reference value of the current time point is corrected, and a first corrected water adding amount reference value of the current time point is obtained; according to the first corrected water adding amount reference value, water is added at a leaf moistening and feeding inlet; an outlet moisture value corresponding to the first corrected water adding amount reference value is collected, and an error is formed; the first corrected water adding amount reference value, the error and the error change rate are input into a fuzzy PID controller, and the fuzzy PID controller outputs a second corrected water adding amount reference value; and at least one of the steam pressure, the water spraying amount, the air amount and the temperature is adjusted based on the second corrected water adding amount reference value. The water adding precision of the charging machine is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of tobacco leaf moistening and feeding, and in particular, to a method and system for controlling tobacco leaf moistening and feeding, an electronic device, and a computer medium. Background Art

[0002] In the existing tobacco leaf moistening and feeding process, after pre-treating the tobacco leaves, the tobacco leaves are sent into a rotating drum, and humidification and feeding are achieved by spraying steam, water, and liquid materials.

[0003] However, the existing technology only relies on the feedback of the moisture content at the outlet of the tobacco leaves to adjust the water spraying amount, without considering the environmental characteristics, resulting in insufficient control accuracy and frequent manual intervention, increasing the risk of production shutdown. Summary of the Invention

[0004] In view of this, one of the purposes of the embodiments of the present application is to provide a method for controlling tobacco leaf moistening and feeding, which can improve the problem of insufficient control accuracy in the existing technology.

[0005] To achieve the above technical purpose, the technical solution adopted by the present application is as follows:

[0006] In a first aspect, the embodiments of the present application provide a method for controlling tobacco leaf moistening and feeding, and the method includes:

[0007] Collect real-time data, where the real-time data includes the moisture content of the tobacco leaves at the current time point, the instantaneous flow rate of the tobacco leaves at the current time point, and the temperature of the tobacco leaves at the current time point;

[0008] Collect real-time data, where the real-time data includes the moisture content of the tobacco leaves at the current time point, the instantaneous flow rate of the tobacco leaves at the current time point, and the temperature of the tobacco leaves at the current time point;

[0009] Based on the real-time data, according to a preset LSTM model, obtain the water addition amount reference value at the current time point;

[0010] Obtain the environmental parameters at the current time point, and based on a preset correction algorithm, correct the water addition amount reference value at the current time point to obtain the first corrected water addition amount reference value at the current time point;

[0011] Add water at the feeding end of the tobacco leaf moistening machine according to the first corrected water addition amount reference value;

[0012] Collect the outlet moisture value corresponding to the first corrected water addition amount reference value to form an error, and the error is calculated according to the first corrected water addition amount reference value and the outlet moisture value;

[0013] Input the first corrected water addition amount reference value, the error, and the error change rate into a fuzzy PID controller, and the fuzzy PID controller outputs a second corrected water addition amount reference value;

[0014] Based on the second corrected water addition benchmark value, adjust at least one of the steam pressure, water injection volume, air volume, and temperature so that the error between the actual outlet moisture content and the predicted outlet moisture content of the leaf conditioning is less than or equal to a preset threshold.

[0015] Further, obtaining the water addition benchmark value at the current time point based on the real-time data according to a preset LSTM model includes:

[0016] Obtain the theoretical moisture absorption rate at the current time point according to the real-time data;

[0017] Obtain the water addition benchmark value according to the theoretical moisture absorption rate;

[0018] The obtaining the theoretical moisture absorption rate at the current time point according to the real-time data includes:

[0019] Input the real-time data into the LSTM model, and the LSTM model outputs the theoretical moisture absorption rate based on the input real-time data;

[0020] The LSTM model is obtained by training based on historical real-time data and historical theoretical moisture absorption rates. The historical real-time data corresponds to the historical theoretical moisture absorption rates, and the historical real-time data includes historical tobacco leaf moisture content, historical instantaneous tobacco leaf flow rate, and historical tobacco leaf temperature.

[0021] Further, obtaining the environmental parameters at the current time point, and correcting the water addition benchmark value at the current time point based on a preset correction algorithm to obtain the first corrected water addition benchmark value at the current time point includes:

[0022] Obtain the environmental parameters, where the environmental parameters include the temperature at the current time point, the humidity at the current time point, and the seasonal factor at the current time point;

[0023] Obtain the first corrected water addition benchmark value according to the environmental parameters and the water addition benchmark value;

[0024] The obtaining the first corrected water addition benchmark value according to the environmental parameters and the water addition benchmark value includes:

[0025] Q 修正1 = Q 基准 ×(1 + α·ΔT + β·ΔH + γ·S)

[0026] ΔH = k H (H 实际 - H 标准 )

[0027] Wherein, Q 修正1 represents the first corrected water addition benchmark value;

[0028] Q 基准represents the reference value of the water addition amount;

[0029] ΔT represents the difference between the temperature at the current time point and the preset reference temperature;

[0030] ΔH represents the humidity deviation coefficient;

[0031] k H represents the humidity sensitivity coefficient;

[0032] H 实际 represents the humidity at the current time point;

[0033] H 标准 represents the preset reference humidity;

[0034] S is the season factor, taking -0.05 to -0.08 in the plum rain season and 0.03 to 0.05 in the dry season;

[0035] α, β, and γ respectively represent the weight coefficients of the temperature difference, humidity deviation, and season factor.

[0036] Further, inputting the first corrected reference value of the water addition amount, the error, and the error change rate into the fuzzy PID controller, and the fuzzy PID controller outputs the second corrected reference value of the water addition amount, including:

[0037] The fuzzy PID controller adjusts the PID parameters based on the error and the error change rate, and the PID parameters include the proportional coefficient, integral coefficient, and differential coefficient;

[0038] Based on the adjusted PID parameters and the first corrected reference value of the water addition amount, obtain the second corrected reference value of the water addition amount;

[0039] The fuzzy PID controller adjusts the PID parameters based on the error and the error change rate, including:

[0040] When the errors collected at at least two time points indicate that the error increases, and the error change rates collected at at least two time points also increase, increase the proportional coefficient;

[0041] When the error change rate of the error collected at at least two time points is less than the preset change rate, increase the integral coefficient;

[0042] When the errors collected at at least two time points indicate that the error decreases, and the decreasing amplitude is greater than the preset amplitude, increase the differential coefficient.

[0043] Further, adjusting at least one of the steam pressure, water injection amount, hot air volume, and hot air temperature based on the second corrected reference value of the water addition amount, including:

[0044] When the second corrected water addition reference values collected at at least two time points indicate an increase in the second corrected water addition reference value, increase or decrease the steam pressure;

[0045] Determine the water injection amount based on the second corrected water addition reference value;

[0046] When the second corrected water addition reference values collected at at least two time points indicate an increase in the second corrected water addition reference value, increase or decrease the hot air volume;

[0047] and / or

[0048] When the second corrected water addition reference values collected at at least two time points indicate an increase in the second corrected water addition reference value, increase or decrease the hot air temperature.

[0049] Furthermore, the determining the water injection amount based on the second corrected water addition reference value includes:

[0050] Q spray =Q 修正2 ×F in ×(1 + ΔT·k T )

[0051] wherein, Q spray represents the water injection amount;

[0052] Q 修正2 represents the second corrected water addition reference value;

[0053] F in represents the instantaneous tobacco leaf flow rate;

[0054] ΔT represents the difference between the temperature at the current time point and the preset reference temperature;

[0055] k T represents the temperature compensation coefficient, which is 0.02 - 0.05.

[0056] In a second aspect, an embodiment of the present application provides a tobacco leaf moistening and feeding control system, including:

[0057] A first acquisition module configured to acquire real-time data, where the real-time data includes the moisture content of tobacco leaves at the current time point, the instantaneous tobacco leaf flow rate at the current time point, and the temperature of tobacco leaves at the current time point;

[0058] A first calculation module configured to obtain the water addition reference value at the current time point based on the real-time data;

[0059] A second calculation module, configured to obtain environmental parameters at the current time point, correct the water addition benchmark value at the current time point, and obtain a first corrected water addition benchmark value at the current time point;

[0060] A first control module, configured to add water at the inlet of the leaf conditioning and feeding according to the first corrected water addition benchmark value;

[0061] A second acquisition module, configured to acquire an outlet moisture value corresponding to the first corrected water addition benchmark value, form an error, and the error is calculated according to the first corrected water addition benchmark value and the outlet moisture value;

[0062] A third calculation module, configured to input the first corrected water addition benchmark value, the error, and the error change rate into a fuzzy PID controller, and the fuzzy PID controller outputs a second corrected water addition benchmark value;

[0063] A second control module, configured to adjust at least one of steam pressure, water injection amount, air volume, and temperature based on the second corrected water addition benchmark value, so that the error between the actual outlet moisture of the leaf conditioning and the expected outlet moisture is less than or equal to a preset threshold.

[0064] In a third aspect, an embodiment of the present application provides an electronic device, which includes a processor and a memory coupled to each other. The memory stores a computer program. When the computer program is executed by the processor, the electronic device executes the above method.

[0065] In a third aspect, an embodiment of the present application provides a computer-readable storage medium, in which a computer program is stored. When the computer program runs on a computer, the computer executes the above method.

[0066] The invention adopting the above technical solution has the following advantages:

[0067] In the technical solution provided by the present application, the water addition benchmark value at the current time point is calculated by collecting real-time data, and then the water addition benchmark value is corrected according to environmental parameters to obtain a first corrected water addition benchmark value. After adding water at the inlet of the leaf conditioning and feeding according to the first corrected water addition benchmark value, the first corrected water addition benchmark value is further corrected by a fuzzy PID controller to obtain a second corrected water addition benchmark value. Finally, according to the second corrected water addition benchmark value, the steam pressure, water injection amount, air volume, and temperature are adjusted. This solution combines environmental parameters, takes into account the influence of the environment on the water addition amount, corrects the water addition benchmark value according to environmental parameters, improves the water addition accuracy of the feeder, and at the same time, through the fuzzy PID controller, adjusts parameters such as steam, air volume, and temperature in a linkage manner to achieve the energy-to-tobacco balance in the leaf conditioning process and reduce the problems of tobacco leaf caking or over-drying caused by process fluctuations. Brief Description of the Drawings

[0068] This application can be further illustrated by the non-limiting embodiments shown in the drawings. It should be understood that the following drawings only show some embodiments of this application, and thus should not be regarded as limiting the scope. For those of ordinary skill in the art, without creative efforts, other relevant drawings can also be obtained based on these drawings.

[0069] Figure 1 It is a flowchart of S110 - S170 provided by an embodiment of this application.

[0070] Figure 2 Provided by an embodiment of this application Figure 1 Sub - flowchart of S120

[0071] Figure 3 Provided by an embodiment of this application Figure 1 Sub - flowchart of S130

[0072] Figure 4 Provided by an embodiment of this application Figure 1 Sub - flowchart of S150

[0073] Figure 5 It is a block diagram of the system provided by an embodiment of this application.

[0074] Icons: 1 - First acquisition module; 2 - First calculation module; 3 - Second calculation module; 4 - First control module; 5 - Second acquisition module; 6 - Third calculation module; 7 - Second control module. Detailed Description of the Embodiments

[0075] The following will describe this application in detail in combination with the drawings and specific embodiments. It should be noted that in the drawings or the description of the specification, similar or identical parts use the same reference numerals. The implementation manners not shown or described in the drawings are in the forms known to those of ordinary skill in the art. In the description of this application, terms such as "first", "second", etc. are only used for differential description and cannot be understood as indicating or implying relative importance.

[0076] An embodiment of this application provides an electronic device that may include a processing module and a storage module. A computer program is stored in the storage module. When the computer program is executed by the processing module, the electronic device can execute the corresponding steps in the following leaf - feeding control method.

[0077] In this embodiment, the processing module may be an integrated circuit chip with signal processing capabilities. The above-mentioned processing module may be a general-purpose processor. For example, the processor may be a Central Processing Unit (CPU), a Digital Signal Processor (DSP), an Application Specific Integrated Circuit (ASIC), a Field-Programmable Gate Array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, which can implement or execute the various methods, steps, and logic block diagrams disclosed in the embodiments of the present application.

[0078] The storage module may be, but is not limited to, a random access memory, a read-only memory, a programmable read-only memory, an erasable programmable read-only memory, an electrically erasable programmable read-only memory, etc. In this embodiment, the storage module may be used to store preset numbers, etc. Of course, the storage module may also be used to store programs, and after receiving an execution instruction, the processing module executes the program.

[0079] Please refer to Figure 1 , the present application also provides a method for controlling leaf moistening and feeding. Among them, the method for controlling leaf moistening and feeding may include the following steps:

[0080] S110, collect real-time data, where the real-time data includes the moisture content of tobacco leaves at the current time point, the instantaneous flow rate of tobacco leaves at the current time point, and the temperature of tobacco leaves at the current time point;

[0081] S120, based on the real-time data, according to the preset LSTM model, obtain the water addition reference value at the current time point;

[0082] S130, obtain the environmental parameters at the current time point, and based on the preset correction algorithm, correct the water addition reference value at the current time point to obtain the first corrected water addition reference value at the current time point;

[0083] S140, add water at the inlet of leaf moistening and feeding according to the first corrected water addition reference value;

[0084] S150, collect the outlet moisture value corresponding to the first corrected water addition reference value to form an error, and the error is calculated according to the first corrected water addition reference value and the outlet moisture value;

[0085] S160, input the first corrected water addition reference value, the error, and the error change rate into a fuzzy PID controller, and the fuzzy PID controller outputs the second corrected water addition reference value;

[0086] S170, based on the second corrected water addition reference value, adjust at least one of the steam pressure, water injection volume, air volume, and temperature so that the error between the actual outlet moisture content and the predicted outlet moisture content of the leaf conditioning is less than or equal to a preset threshold value.

[0087] The following will elaborate on each step of the leaf conditioning and feeding control method in detail as follows:

[0088] In S110, exemplarily, real-time data can be collected in the following manner, specifically including:

[0089] I. Collection of tobacco leaf moisture content

[0090] Online monitoring with an infrared moisture meter uses non-contact infrared spectroscopy technology to calculate the moisture content based on the absorption characteristics of tobacco leaves for infrared light of a specific wavelength, with an accuracy of up to ±0.1% (is there an extra 4?). Deployment location: Usually installed at the inlet or outlet of the leaf conditioner, sampling synchronously with the tobacco leaf conveyance. The microwave moisture sensor measures moisture based on the change in the dielectric constant when microwaves penetrate the tobacco leaves and is applicable to high-humidity environments (such as the loose re-damping process)

[0091] II. Collection of instantaneous tobacco leaf flow rate

[0092] Dynamic metering with an electronic belt scale measures the weight of tobacco leaves per unit length of the belt through a load cell and calculates the instantaneous flow rate in combination with a speed sensor (formula: Q = q × v), with an accuracy of ±1 kg / h. Non-contact measurement with a nuclear scale calculates the flow rate using the attenuation characteristics of γ-rays when penetrating the tobacco leaves and is applicable to high-temperature and high-dust environments.

[0093] III. Collection of tobacco leaf temperature

[0094] Infrared thermometer, non-contact measurement of the surface temperature of tobacco leaves, response time ≤ 0.5 seconds, temperature measurement range -20 to 150 °C, accuracy ±0.5 °C.

[0095] In S120, as Figure 2 shown, S120 includes the following steps:

[0096] S121: Obtain the theoretical moisture absorption rate at the current time point based on the real-time data;

[0097] S122: Obtain the water addition reference value based on the theoretical moisture absorption rate.

[0098] In S121, the real-time data is input into the LSTM model. Based on the input real-time data, the LSTM model outputs the theoretical moisture absorption rate. The LSTM model is trained based on historical real-time data and historical theoretical moisture absorption rates. The historical real-time data corresponds to the historical theoretical moisture absorption rates. The historical real-time data includes historical tobacco leaf moisture content, historical instantaneous tobacco leaf flow rate, and historical tobacco leaf temperature.

[0099] The training method of the LSTM model in this embodiment can be as follows:

[0100] Method 1: Multivariate time series supervised learning (single-step prediction)

[0101] 1. Data preprocessing and feature engineering

[0102] Feature normalization: According to the following formula, perform Min-Max normalization (range 0-1) on historical tobacco leaf moisture content, tobacco leaf flow rate, and temperature to eliminate the difference in dimensions.

[0103]

[0104] X: The original data values of historical tobacco leaf moisture content, historical tobacco leaf flow rate, and historical temperature;

[0105] X min represents the preset minimum value, such as the preset minimum values of historical tobacco leaf moisture content, historical tobacco leaf flow rate, and historical temperature;

[0106] X max represents the preset maximum value, such as the preset maximum values of historical tobacco leaf moisture content, historical tobacco leaf flow rate, and historical temperature.

[0107] X norm is the normalized standardized value.

[0108] Sequence construction: Generate input-output pairs using a sliding window. The window length is recommended to be 20-50 time steps (to be determined by the autocorrelation function);

[0109] Input sequence: Moisture content + flow rate + temperature (three-dimensional features) at [t-19, t-18,..., t] moments;

[0110] Output sequence: Theoretical moisture absorption rate at t+1 moment (single target);

[0111] Missing value processing: Use linear interpolation to fill in missing data and filter outliers using the 3σ principle.

[0112] 2. LSTM network architecture design

[0113] Python, based on the following code:

[0114] model = Sequential()

[0115] # Input layer: three-dimensional features (moisture content + flow rate + temperature), time step = window length

[0116] model.add(LSTM(units = 64, input_shape = (window_size, 3), return_sequences = True))model.add(Dropout(0.2))model.add(LSTM(units = 32))

[0117] model.add(Dense(1))

[0118] # Output the theoretical moisture absorption rate

[0119] Parameter configuration: Activation function: tanh for hidden layers, linear activation for output layer. Loss function: Mean Squared Error (MSE). Optimizer: Adam (learning rate 0.001, β1 = 0.9, β2 = 0.999)

[0120] 3. Training strategy

[0121] Cross-validation: Use TimeSeriesSplit for time series cross-validation to prevent data leakage. Early stopping mechanism: Monitor the validation set loss and terminate training if there is no improvement for 5 consecutive epochs. Batch training: The batch size is recommended to be 32 - 128, and the number of epochs is set to 50 - 200 (needs to be combined with learning rate decay)

[0122] 4. Application scenarios. Suitable for real-time control scenarios, which require quick response to the latest working condition changes. Case of a cigarette factory: window length = 30, prediction error reduced from ±1.8% to ±0.6%

[0123] Method 2: Encoder-Decoder (Seq2Seq) structure (multi-step prediction)

[0124] 1. Data reconstruction

[0125] Input sequence: Three-dimensional features (moisture content + flow rate + temperature) of the past 50 - 100 time steps. Output sequence: Theoretical moisture absorption rate sequence of the next 5 - 10 time steps (multi-step prediction);

[0126] Data augmentation: Increase sample diversity through TimeWarping;

[0127] 2. Model architecture

[0128] Python, based on the following code

[0129] # Encoder

[0130] # Define the encoder input layer: The input is a sequence with variable time steps, and each time step contains 3 features (such as multi-dimensional sensor data). encoder_inputs = Input(shape=(None, 3)) # Input dimension [batch_size, timesteps, 3]

[0131] # Define the LSTM layer: 128 neurons, return the final state (hidden state h and cell state c)

[0132] encoder_lstm = LSTM(128, return_state=True) # return_state=True indicates returning the final state of the LSTM

[0133] # Run the encoder LSTM: encoder_outputs is the output vector of the last time step (not used), state_h and state_c are the hidden state and cell state

[0134] encoder_outputs, state_h, state_c = encoder_lstm(encoder_inputs) # encoder_outputs: [batch_size, 128] # Save the final state of the encoder as a list for decoder initialization. encoder_states = [state_h, state_c] # State transfer mechanism

[0135] # Decoder

[0136] # Define the decoder input layer: The input is a sequence with variable time steps, and each time step contains 1 feature (such as single-variable prediction). decoder_inputs = Input(shape=(None, 1)) # Input dimension [batch_size, timesteps, 1] # Define the decoder LSTM: 128 neurons, return the complete output sequence (not only the last time step), and receive the encoder state as the initial state

[0137] decoder_lstm = LSTM(128, return_sequences=True, return_state=True) # return_sequences=True retains the output of all time steps#

[0138] Run the decoder LSTM: The initial state comes from the encoder

[0139] decoder_outputs, _, _ = decoder_lstm(decoder_inputs, initial_state=encoder_states) # decoder_outputs: [batch_size, timesteps, 128]

[0140] # Define the fully connected layer: map the LSTM output to the target dimension (single variable prediction)

[0141] decoder_dense = Dense(1) # Output dimension [batch_size, timesteps, 1]

[0142] # Apply the fully connected layer to generate the final prediction result

[0143] decoder_outputs = decoder_dense(decoder_outputs) # Output shape aligns with the target sequence. Attention mechanism: Bahdanau attention layer can be added to improve the prediction accuracy of long sequences.

[0144] 3. Training optimization

[0145] Curriculum learning: First train single-step prediction and gradually increase the output step size

[0146] Hybrid loss function:

[0147] Loss = 0.7·MSE + 0.3·MAE

[0148] Regularization: Weight L2 regularization (λ = 0.001) + Gradient clipping (threshold = 1.0).

[0149] The application scenario is suitable for process optimization scenarios, and it is necessary to predict the moisture absorption rate trend at multiple future time points. A certain case: Predict the moisture absorption rate in the next 10 minutes, and the steam regulation response speed is increased by 40%.

[0150] In S122, calculate the reference value of the water addition amount in combination with the mass conservation equation. The specific formula is as follows:

[0151]

[0152] Among them, H in is the inlet moisture content, m is the inlet tobacco leaf flow rate, t res is the tobacco leaf residence time, ΔH is the theoretical moisture absorption rate, Q 0 is the reference value of the water addition amount, H target is the target outlet moisture content.

[0153] In S130, if Figure 3As shown, S130 mainly includes the following steps:

[0154] S131: Obtain the environmental parameters, where the environmental parameters include the temperature at the current time point, the humidity at the current time point, and the season factor at the current time point;

[0155] S132: Obtain the first corrected water addition benchmark value according to the environmental parameters and the water addition benchmark value. The first corrected water addition benchmark value is obtained according to the following formula:

[0156] Q 修正1 =Q 基准 ×(1 + α·ΔT + β·ΔH + γ·S

[0157] ΔH=k H (H 实际 -H 标准 )

[0158] Where, Q 修正1 represents the first corrected water addition benchmark value;

[0159] Q 基准 represents the water addition benchmark value;

[0160] ΔT represents the difference between the temperature at the current time point and the preset reference temperature;

[0161] ΔH represents the humidity deviation coefficient;

[0162] k H represents the humidity sensitivity coefficient;

[0163] H 实际 represents the humidity at the current time point;

[0164] H 标准 represents the preset reference humidity;

[0165] S is the season factor, taking -0.05 to -0.08 in the plum rain season and 0.03 to 0.05 in the dry season;

[0166] α, β, and γ respectively represent the weight coefficients of the temperature difference, humidity deviation, and season factor.

[0167] In step 140, water can be added at the inlet of the leaf conditioning and feeding in the following two ways. Specifically:

[0168] Method 1: Dynamic adjustment of steam atomization by a dual-medium nozzle

[0169] The technical principle uses a steam-water dual-medium nozzle to atomize liquid water into micron-sized particles (atomization particle size ≤ 50 μm) by high-pressure steam and uniformly spray them onto the surface of the tobacco leaves.

[0170] Steam pressure and water addition amount linkage control: Steam pressure regulation: Dynamically adjust the steam pressure according to the reference value (typical range 0.2 - 0.6 MPa). The higher the pressure, the finer the atomization effect, but it is necessary to avoid excessive vaporization resulting in water loss.

[0171] Water amount dynamic matching: Adjust the water spraying amount in real time through a proportional valve to ensure that the atomized water amount is consistent with the reference value (error ≤ ±1.0%).

[0172] Control key points

[0173] Atomization angle optimization: The installation angle of the nozzle needs to match the rotation direction of the drum (recommended inclination angle 30° - 45°), covering the rolling track of the tobacco leaves to avoid local over-wetting or drying.

[0174] Hot air assisted penetration: Simultaneously turn on the hot air circulation system (temperature 45 - 55 °C) to accelerate the penetration of the atomized water and reduce the evaporation loss of the surface residual water.

[0175] Method 2: Control the valve size through a valve positioner for precise flow control

[0176] Technical principle: Adopt a high-precision valve positioner, such as a magnetic drive gear pump, and directly adjust the water addition amount by controlling the opening of the valve positioner through PID closed-loop control.

[0177] Flow feedback compensation: Real-time collect the actual flow signal (such as the data of an electromagnetic flowmeter), compare it with the reference value to generate an error signal; the PID controller outputs a rotational speed correction instruction (response time ≤ 0.5 s) to eliminate the instantaneous flow deviation.

[0178] Control algorithm optimization

[0179] Segmented PID parameters: Switch the control parameters according to the flow range (such as increasing the integral effect at low flow to prevent overshoot);

[0180] Feedforward compensation: Introduce the instantaneous flow of the tobacco leaves as a feedforward signal to predict the impact of flow mutation on the water addition amount (such as when the flow suddenly increases by 10%, increase the opening by 5% in advance).

[0181] In S150, the training method of the fuzzy PID controller can be:

[0182] I. Offline training (supervised learning) based on historical data

[0183] 1. Data preparation and feature engineering

[0184] Input features:

[0185] The first corrected water addition amount reference value (Q_adj1)

[0186] Outlet moisture error (e = target moisture - actual moisture)

[0187] Rate of change of error (Δe = e(t) - e(t - 1)

[0188] Output label: Second corrected water addition reference value (Q_adj2) or PID parameter increment (ΔKp, ΔKi, ΔKd) Data cleaning:

[0189] Eliminate sensor fault data (such as instantaneous flow rate being 0 or over - limit value);

[0190] Align timestamps (ensure input - output time window synchronization).

[0191] 2. Fuzzy rule base construction

[0192] Expert experience rules:

[0193] Formulate initial rules according to the knowledge of process engineers (for example: "If the error is positive large and the rate of change of error is negative small, then increase the proportional coefficient Kp");

[0194] Rule optimization: Screen effective rule combinations through genetic algorithm (GA) or particle swarm optimization (PSO), and eliminate redundant rules (the number of rules is compressed from 49 to 25);

[0195] 3. Membership function optimization

[0196] Fuzzification of input variables:

[0197] The membership functions of error (e) and rate of change of error (Δe) adopt Gaussian or triangular distributions, and adjust the center point and width through gradient descent.

[0198] Defuzzification of output variables:

[0199] Adopt the centroid method to generate the precise control quantity, and optimize the defuzzification weight through backpropagation.

[0200] 4. Training process

[0201] python# Example: Training framework of fuzzy PID controller based on Keras

[0202] from pyFTS.pid import FuzzyPID

[0203] from sklearn.model_selection import train_test_split

[0204] # Load historical data set (input: Q_adj1, e, Δe; output: Q_adj2)

[0205] X_train, X_test, y_train, y_test = train_test_split(data_inputs, data_labels, test_size = 0.2)

[0206] # Initialize the fuzzy PID controller

[0207] fp = FuzzyPID(n_rules = 25, input_mf = 'gaussian', output_defuzz = 'centroid')

[0208] # Optimize the rule base using the genetic algorithm

[0209] fp.train_GA(X_train, y_train, population_size = 50, generations = 100)

[0210] # Verify the model performance

[0211] mae = fp.evaluate(X_test, y_test) print(f"MAE of the test set: {mae:.2f} L / min")

[0212] 5. Offline verification

[0213] Digital twin test:

[0214] Inject historical data into the virtual leaf conditioning machine model to verify the control stability (e.g., the outlet moisture fluctuation ≤ ±0.5%).

[0215] Robustness analysis: Simulate sensor noise (±0.2% moisture content error) and execution delay (0.5 - 2 seconds), and evaluate the overshoot and oscillation amplitude of the control quantity.

[0216] II. Online adaptive training (reinforcement learning)

[0217] 1. Definition of the state-action space

[0218] State:

[0219] Current error (e_t), integral of error (∑e), differential of error (Δe)

[0220] Environmental parameters (temperature, humidity);

[0221] Action:

[0222] PID parameter adjustment amount (ΔKp, ΔKi, ΔKd) or directly output Q_adj2

[0223] Reward function: R = -(α∣e∣ + β∣Δe∣ + γ∣ΔQadj2∣), where α = 0.6 (error penalty), β = 0.3 (fluctuation penalty), and γ = 0.1 (control variable change penalty).

[0224] 2. Algorithm Selection and Training

[0225] Deep Deterministic Policy Gradient (DDPG):

[0226] The Actor network outputs the PID parameter adjustment strategy;

[0227] The Critic network evaluates the state-action value;

[0228] Training steps:

[0229] Initialize the Actor-Critic network parameters;

[0230] Real-time collect state data (e, Δe, T, RH);

[0231] Execute actions (adjust Kp / Ki / Kd) and observe the change in outlet moisture;

[0232] Calculate the reward value and update the network weights;

[0233] Save the model snapshot every 1 hour to prevent system instability;

[0234] 3. Online Safety Mechanism

[0235] Action constraint: Limit the adjustment range of PID parameters (e.g., single-step ΔKp ≤ ±10%);

[0236] Emergency fallback: When the reward value is lower than the threshold for 5 consecutive times, switch to the backup PID controller;

[0237] Prioritized experience replay: Store high-reward value samples to accelerate policy convergence.

[0238] In this embodiment, as Figure 4 shown, S150 includes the following steps:

[0239] S151: The fuzzy PID controller adjusts the PID parameters based on the error and the error change rate, and the PID parameters include the proportional coefficient, the integral coefficient, and the differential coefficient;

[0240] S152: Based on the adjusted PID parameters and the first corrected water addition reference value, obtain the second corrected water addition reference value.

[0241] In S151, based on the adjusted PID parameters and the first corrected water addition reference value, obtain the second corrected water addition reference value;

[0242] The fuzzy PID controller adjusts the PID parameters based on the error and the error change rate, including:

[0243] When the errors collected at at least two time points indicate that the error increases, and the error change rates collected at at least two time points also increase, increase the proportional coefficient (Kp). When the outlet moisture suddenly deviates from the target value, increasing Kp can accelerate the response;

[0244] When the error change rate of the error collected at at least two time points is less than the preset change rate, increase the integral coefficient (K i ). When the long-term environmental temperature and humidity fluctuations cause the cumulative error of the water addition amount, the dynamic adjustment of K i can avoid over-integration.

[0245] When the errors collected at at least two time points indicate that the error decreases, and the decreasing amplitude is greater than the preset amplitude, increase the differential coefficient. When the sudden change of steam pressure causes the rapid change of the outlet moisture, the adjustment of K d can enhance the system damping.

[0246] In this embodiment, a fuzzy PID controller is used to correct the first corrected water addition amount reference value. The fuzzy PID controller has the following advantages compared with the traditional PID controller:

[0247] 1. Nonlinear adaptability

[0248] Traditional PID relies on fixed parameters and is difficult to cope with nonlinear disturbances such as environmental temperature and humidity and tobacco leaf flow rate during leaf moistening and feeding.

[0249] 2. The fuzzy PID can adapt to extreme working conditions of high humidity (compensation coefficient -0.07) in the rainy season and low humidity (compensation coefficient +0.04) in the dry season through dynamic parameter adjustment.

[0250] 3. Anti-interference and robustness

[0251] The fuzzy rule base can embed process knowledge (such as "freezing compensation coefficient when steam pressure deviation > 10%") to avoid out-of-control caused by parameter mutation.

[0252] In S160, the ways to adjust the steam pressure, water injection amount, hot air volume and hot air temperature include:

[0253] 1. When the second corrected water addition amount reference value collected at at least two time points indicates that the second corrected water addition amount reference value increases, increase or decrease the steam pressure.

[0254] 2. Determine the water injection amount based on the second corrected water addition amount reference value. The specific formula is:

[0255] Q spray =Q 修正2 ×F in ×(1+ΔT·k T )

[0256] Among them, Q spray Indicates the amount of water sprayed;

[0257] Q 修正2 Indicates the reference value of water addition after the second correction;

[0258] F in Indicates the instantaneous flow rate of the tobacco leaves;

[0259] ΔT represents the difference between the temperature at the current time point and the preset reference temperature;

[0260] k T Indicates the temperature compensation coefficient, which is 0.02-0.05.

[0261] 3. When the second corrected water addition amount reference value collected at at least two time points indicates that the second corrected water addition amount reference value increases, increase or decrease the hot air volume;

[0262] 4. When the second corrected water addition amount reference value collected at at least two time points indicates that the second corrected water addition amount reference value increases, increase or decrease the hot air temperature.

[0263] The present application also provides a leaf moistening feeding control system, which includes at least one software function module that can be stored in a storage module or fixed in an operating system (OS) in the form of software or firmware. The processing module is used to execute the executable module stored in the storage module, such as the software function module and computer program included in the leaf moistening feeding control system.

[0264] like Figure 5 As shown, the functions of each module of the Runye feeding control system can be as follows:

[0265] A first acquisition module 1 is configured to acquire real-time data, wherein the real-time data includes the moisture content of tobacco leaves at a current time point, the instantaneous flow rate of tobacco leaves at a current time point, and the temperature of tobacco leaves at a current time point;

[0266] The first calculation module 2 is configured to obtain a water addition amount reference value at a current time point based on real-time data and a preset LSTM model;

[0267] The second calculation module 3 is configured to obtain the environmental parameters at the current time point, correct the water addition amount reference value at the current time point, and obtain a first corrected water addition amount reference value at the current time point;

[0268] The first control module 4 is configured to add water at the inlet of the leaf conditioning and feeding according to the first corrected water addition reference value.

[0269] The second acquisition module 5 is configured to acquire the outlet moisture value corresponding to the first corrected water addition reference value to form an error, and the error is calculated based on the first corrected water addition reference value and the outlet moisture value.

[0270] The third calculation module 6 is configured to input the first corrected water addition reference value, the error, and the error change rate into a fuzzy PID controller, and the fuzzy PID controller outputs a second corrected water addition reference value.

[0271] The second control module 7 is configured to adjust at least one of the steam pressure, the water injection amount, the air volume, and the temperature based on the second corrected water addition reference value, so that the error between the actual outlet moisture of the leaf conditioning and the expected outlet moisture is less than or equal to a preset threshold.

[0272] Among them, the first calculation module 2 obtains the water addition reference value based on the following algorithm. It includes:

[0273] Based on the real-time data, obtain the theoretical moisture absorption rate at the current time point;

[0274] Based on the theoretical moisture absorption rate, obtain the water addition reference value;

[0275] The obtaining of the theoretical moisture absorption rate at the current time point based on the real-time data includes:

[0276] Input the real-time data into an LSTM model, and the LSTM model outputs the theoretical moisture absorption rate based on the input real-time data;

[0277] The LSTM model is obtained by training based on historical real-time data and historical theoretical moisture absorption rates. The historical real-time data corresponds to the historical theoretical moisture absorption rates, and the historical real-time data includes historical tobacco leaf moisture content, historical instantaneous tobacco leaf flow rate, and historical tobacco leaf temperature.

[0278] The second calculation module 3 obtains the first corrected water addition reference value based on the following algorithm, including:

[0279] Obtain the environmental parameters, where the environmental parameters include the temperature at the current time point, the humidity at the current time point, and the seasonal factor at the current time point;

[0280] Based on the environmental parameters and the water addition reference value, obtain the first corrected water addition reference value;

[0281] The obtaining of the first corrected water addition reference value based on the environmental parameters and the water addition reference value includes:

[0282] Q 修正1 =Q 基准 ×(1 + α·ΔT + β·ΔH + γ·S

[0283] ΔH = k H (H 实际 -H 标准 )

[0284] Among them, Q 修正1 represents the first corrected water addition benchmark value;

[0285] Q 基准 represents the water addition benchmark value;

[0286] ΔT represents the difference between the temperature at the current time point and the preset reference temperature;

[0287] ΔH represents the humidity deviation coefficient;

[0288] k H represents the humidity sensitivity coefficient;

[0289] H 实际 represents the humidity at the current time point;

[0290] H 标准 represents the preset reference humidity;

[0291] S is the season factor, taking -0.05 to -0.08 in the rainy season and 0.03 to 0.05 in the dry season;

[0292] α, β, and γ respectively represent the weight coefficients of the temperature difference, humidity deviation, and season factor.

[0293] The third calculation module 6 calculates the second corrected water addition benchmark value according to the following method:

[0294] Inputting the first corrected water addition benchmark value, error, and error change rate into the fuzzy PID controller, and the fuzzy PID controller outputs the second corrected water addition benchmark value, including:

[0295] The fuzzy PID controller adjusts the PID parameters based on the error and error change rate, and the PID parameters include the proportional coefficient, integral coefficient, and differential coefficient;

[0296] Based on the adjusted PID parameters and the first corrected water addition benchmark value, the second corrected water addition benchmark value is obtained;

[0297] The fuzzy PID controller adjusts the PID parameters based on the error and error change rate, including:

[0298] When the errors collected at at least two time points indicate an increase in the error, and the error change rates collected at at least two time points also increase, increase the proportionality coefficient;

[0299] When the error change rate of the error collected at at least two time points is less than the preset change rate, increase the integral coefficient;

[0300] When the errors collected at at least two time points indicate a decrease in the error, and the decreasing amplitude is greater than the preset amplitude, increase the differential coefficient.

[0301] The second control module 7 adjusts the steam pressure, water injection amount, air volume, and temperature based on the following methods:

[0302] When the second corrected water injection amount reference value collected at at least two time points indicates an increase in the second corrected water injection amount reference value, increase or decrease the steam pressure;

[0303] Determine the water injection amount based on the second corrected water injection amount reference value;

[0304] When the second corrected water injection amount reference value collected at at least two time points indicates an increase in the second corrected water injection amount reference value, increase or decrease the hot air volume;

[0305] And / or

[0306] When the second corrected water injection amount reference value collected at at least two time points indicates an increase in the second corrected water injection amount reference value, increase or decrease the hot air temperature.

[0307] Determining the water injection amount based on the second corrected water injection amount reference value includes:

[0308] Q spray = Q 修正2 ×F in ×(1 + ΔT·k T )

[0309] Wherein, Q spray represents the water injection amount;

[0310] Q 修正2 represents the second corrected water injection amount reference value;

[0311] F in represents the instantaneous tobacco leaf flow rate;

[0312] ΔT represents the difference between the temperature at the current time point and the preset reference temperature;

[0313] k T represents the temperature compensation coefficient, which is 0.02 - 0.05.

[0314] It should be noted that those skilled in the art can clearly understand that for the convenience and conciseness of description, the specific working process of the above-described electronic device can refer to the corresponding processes of the steps in the foregoing method, and will not be elaborated herein.

[0315] The embodiment of the present application also provides a computer-readable storage medium. A computer program is stored in the computer-readable storage medium. When the computer program runs on a computer, the computer is enabled to execute the leaf moistening and feeding control method as described in the above embodiment.

[0316] Through the description of the above embodiments, those skilled in the art can clearly understand that the present application can be implemented by hardware, or can be implemented by means of software plus a necessary general hardware platform. Based on such an understanding, the technical solution of the present application can be embodied in the form of a software product. The software product can be stored in a non-volatile storage medium (which can be a CD-ROM, a USB flash drive, a mobile hard disk, etc.), and includes several instructions for enabling a computer device (which can be a personal computer, an electronic device, or a network device, etc.) to execute the methods described in various implementation scenarios of the present application.

[0317] In the embodiments provided by the present application, it should be understood that the disclosed systems and methods can also be implemented in other ways. The system and method embodiments described above are merely illustrative. For example, the flowcharts and block diagrams in the accompanying drawings show the possible architectures, functions, and operations of systems, methods, and computer program products according to multiple embodiments of the present application. In this regard, each block in the flowchart or block diagram may represent a module, a program segment, or a part of code, and the module, program segment, or part of code includes one or more executable instructions for implementing the specified logical function. It should also be noted that each block in the block diagram and / or flowchart, and the combination of blocks in the block diagram and / or flowchart, can be implemented by a dedicated hardware-based system for performing the specified functions or actions, or can be implemented by a combination of dedicated hardware and computer instructions. In addition, the various functional modules in the embodiments of the present application can be integrated together to form an independent part, or each module can exist separately, or two or more modules can be integrated to form an independent part.

[0318] The above is only the embodiment of the present application and is not used to limit the protection scope of the present application. For those skilled in the art, the present application can have various changes and modifications. Any modification,

[0319] equivalent replacement, improvement, etc. made within the spirit and principle of the present application shall be included in the protection scope of the present application.

Claims

1. A leaf moistening feeding control method, characterized in that: The method comprises: Collecting real-time data, the real-time data including the moisture content of tobacco leaves at the current time point, the instantaneous flow rate of tobacco leaves at the current time point, and the temperature of tobacco leaves at the current time point; Based on real-time data and the preset LSTM model, the water addition benchmark value at the current time point is obtained; Acquire the environmental parameters at the current time point, and based on a preset correction algorithm, correct the water addition amount reference value at the current time point to obtain a first corrected water addition amount reference value at the current time point; According to the first corrected water addition reference value, add water at the feeding end of the leaf moistening machine; Collecting the outlet moisture value corresponding to the first corrected water addition reference value to form an error, wherein the error is calculated based on the first corrected water addition reference value and the outlet moisture value; Inputting the first corrected water addition amount reference value, error and error change rate into a fuzzy PID controller, and the fuzzy PID controller outputs a second corrected water addition amount reference value; Based on the second corrected water addition reference value, at least one of the steam pressure, water spraying amount, air volume and temperature is adjusted so that the error between the actual outlet moisture of the leaves and the expected outlet moisture is less than or equal to a preset threshold.

2. The method according to claim 1, characterized in that Based on the real-time data, according to the preset LSTM model, the reference value of the water addition amount at the current time point is obtained, including: According to the real-time data, a theoretical moisture absorption rate at a current time point is obtained; According to the theoretical moisture absorption rate, the reference value of water addition amount is obtained; The step of obtaining the theoretical moisture absorption rate at the current time point according to the real-time data includes: The real-time data is input into the LSTM model, and the LSTM model outputs the theoretical moisture absorption rate based on the input real-time data; The LSTM model is obtained through training based on historical real-time data and historical theoretical moisture absorption rate, the historical real-time data corresponds to the historical theoretical moisture absorption rate, and the historical real-time data includes historical tobacco leaf moisture content, historical tobacco leaf instantaneous flow rate and historical tobacco leaf temperature.

3. The method according to claim 1, characterized in that The step of obtaining the environmental parameters at the current time point, correcting the water addition reference value at the current time point, and obtaining a first corrected water addition reference value at the current time point based on a preset correction algorithm includes: Acquire the environmental parameters, wherein the environmental parameters include the temperature at the current time point, the humidity at the current time point, and the seasonal factor at the current time point; Obtaining the first corrected water addition amount reference value according to the environmental parameter and the water addition amount reference value; The step of obtaining the first corrected water addition amount reference value according to the environmental parameter and the water addition amount reference value comprises: Q 修正1 =Q 基准 ×(1+α·ΔT+β·ΔH+γ·S) ΔH=k H (H 实际 -H 标准 ) Among them, Q 修正1 Indicates the first corrected water addition amount reference value; Q 基准 Indicates the water addition reference value; ΔT represents the difference between the temperature at the current time point and the preset reference temperature; ΔH represents the humidity deviation coefficient; k H Represents the humidity sensitivity coefficient; H 实际 Indicates the humidity at the current time point; H 标准 Indicates the preset reference humidity; S is the seasonal factor, ranging from -0.05 to -0.08 in the rainy season and 0.03 to 0.05 in the dry season; α, β, and γ represent the weight coefficients of temperature difference, humidity deviation, and seasonal factor, respectively.

4. The method according to claim 1, characterized in that: The step of inputting the first corrected water addition amount reference value, the error and the error change rate into a fuzzy PID controller, and the fuzzy PID controller outputting a second corrected water addition amount reference value, comprises: The fuzzy PID controller adjusts PID parameters based on the error and the error change rate, wherein the PID parameters include a proportional coefficient, an integral coefficient, and a differential coefficient; Based on the adjusted PID parameter and the first corrected water addition amount reference value, a second corrected water addition amount reference value is obtained; The fuzzy PID controller adjusts PID parameters based on the error and the error change rate, including: When the errors collected at at least two time points indicate that the errors are increasing, and the error change rate collected at at least two time points is also increasing, increasing the proportionality coefficient; When the error collected at at least two time points indicates that the rate of change of the error is less than a preset rate of change, increasing the integral coefficient; When the error collected at at least two time points indicates that the error is reduced, and the magnitude of the reduction is greater than a preset magnitude, the differential coefficient is increased.

5. The method according to claim 1, characterized in that: The step of adjusting at least one of steam pressure, water spraying amount, hot air volume and hot air temperature based on the second corrected water addition amount reference value comprises: When the second corrected water addition amount reference value collected at at least two time points indicates that the second corrected water addition amount reference value increases, increasing or decreasing the steam pressure; determining the water spraying amount based on the second corrected water addition amount reference value; When the second corrected water addition amount reference value collected at at least two time points indicates that the second corrected water addition amount reference value increases, increasing or decreasing the hot air volume; and / or When the second corrected water addition amount reference value collected at at least two time points indicates that the second corrected water addition amount reference value increases, the hot air temperature is increased or decreased.

6. The method according to claim 5, characterized in that The step of determining the water spraying amount based on the second corrected water addition amount reference value comprises: Q spray =Q 修正2 ×F in ×(1+ΔT·k T ) Among them, Q spray Indicates the amount of water sprayed; Q 修正2 Indicates the reference value of water addition after the second correction; F in Indicates the instantaneous flow rate of the tobacco leaves; ΔT represents the difference between the temperature at the current time point and the preset reference temperature; k T Indicates the temperature compensation coefficient, which is 0.02-0.

05.

7. A leaf moistening and feeding control system, characterized in that: include: A first acquisition module is configured to acquire real-time data, wherein the real-time data includes a tobacco leaf moisture content at a current time point, a tobacco leaf instantaneous flow rate at a current time point, and a tobacco leaf temperature at a current time point; A first calculation module is configured to obtain a water addition reference value at a current time point based on real-time data; A second calculation module is configured to obtain the environmental parameters at the current time point, correct the water addition amount reference value at the current time point, and obtain a first corrected water addition amount reference value at the current time point; A first control module is configured to add water at the inlet of the leaf moistening feed according to the first corrected water addition reference value; A second acquisition module is configured to acquire an outlet moisture value corresponding to the first corrected water addition reference value to form an error, wherein the error is calculated based on the first corrected water addition reference value and the outlet moisture value; A third calculation module is configured to input the first corrected water addition amount reference value, the error and the error change rate into a fuzzy PID controller, and the fuzzy PID controller outputs a second corrected water addition amount reference value; The second control module is configured to adjust at least one of the steam pressure, water spray volume, air volume and temperature based on the second corrected water addition reference value so that the error between the actual outlet moisture of the leaves and the expected outlet moisture is less than or equal to a preset threshold.

8. An electronic device, characterized in that: The electronic device comprises a processor and a memory coupled to each other, wherein the memory stores a computer program. When the computer program is executed by the processor, the electronic device executes the method according to any one of claims 1 to 6.

9. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, and when the computer program is executed on a computer, the computer is enabled to execute the method according to any one of claims 1 to 6.

Images