Self-adaptive control method and system for improving tea oil extraction efficiency

Abstract

The invention discloses a self-adaptive control method for improving tea oil extraction efficiency, which comprises the following steps: carrying out data preprocessing according to historically collected tea seed raw material multi-dimensional data, and further generating a data set containing training data and test data; a neural network is adopted to continuously train the data set, and a corrected control prediction model is generated; deploying a control prediction model, inputting the collected real-time multi-dimensional data of the tea seed raw materials into the control prediction model, and generating tea oil extraction prediction data; and generating a matched operation strategy of the tea oil extraction equipment according to the tea oil extraction prediction data, and dynamically controlling the matched operation parameters of the tea oil extraction equipment according to the matched time interval according to the operation strategy. According to the self-adaptive control method for improving the tea oil extraction efficiency, the tea oil extraction process can be automatically, intelligently and dynamically adjusted, so that the tea oil extraction efficiency is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of tea oil extraction, and in particular to an adaptive control method and system for improving tea oil extraction efficiency. Background Art

[0002] Tea oil is a vegetable oil extracted from tea seeds. It is rich in nutrients and has a unique flavor. Its main component is oil, which contains monounsaturated fatty acids, vitamin E, phytosterols, etc. It has good health benefits. Tea oil is widely used in food, cosmetics and medicine.

[0003] At present, tea oil is usually extracted from tea seeds by a pressing method, in which oil is directly extracted from tea seeds by mechanical pressing such as a hydraulic press or a screw machine. The process is simple and easy to operate. It is easy to operate, green and environmentally friendly, does not use organic solvents, has good quality for the extracted tea oil, and can maintain the natural flavor of the oil. However, the mechanical pressing method has a low extraction rate, usually only 30%-50% of the oil can be extracted, and the residual oil rate is high. Multiple pressing or combination with other extraction methods can improve the oil recovery rate, and it relies on manual control of equipment operating temperature, pressure, time and other parameters in the production process. The degree of automation is low, the process parameters are not adjusted in time, and dynamic adjustments cannot be made according to real-time monitoring data, which affects the overall efficiency of the extraction. Therefore, there is an urgent need for an adaptive control method that can automatically and intelligently adjust the tea oil extraction process to improve the tea oil extraction efficiency. Summary of the invention

[0004] Purpose of the invention: In order to overcome the above shortcomings, the purpose of the present invention is to provide an adaptive control method and system for improving tea oil extraction efficiency.

[0005] In order to solve the above technical problems, the present invention provides an adaptive control method for improving tea oil extraction efficiency, comprising: Step S1: performing data preprocessing based on the multidimensional data of tea seed raw materials collected historically, thereby generating a data set including training data and test data; Step S2: using a neural network to continuously train the data set to generate a revised control prediction model, the training process includes: forward propagating the training data of the data set and transforming it through weighted summation and activation function, calculating the loss function, and using back propagation to update parameters through gradient descent method; Step S3: deploying the control prediction model, and then inputting the collected real-time tea seed raw material multi-dimensional data into the control prediction model to generate tea oil extraction prediction data; Step S4: Generate an operation strategy for a matching tea oil extraction device based on the tea oil extraction prediction data, and then dynamically control the operation parameters of the matching tea oil extraction device according to the operation strategy and in a matching time interval.

[0006] In one aspect, in step S1, the method comprises the following steps: Step S11: performing data cleaning based on historically collected tea seed characteristic data, extraction process parameters, extraction result data and environmental data, wherein the data cleaning includes: outlier and missing value processing; Step S12: standardizing the cleaned data to generate multi-dimensional data of tea seed raw materials; Step S13: dividing the multidimensional data of tea seed raw materials into training data and test data, and then generating a data set including the training data and the test data.

[0007] In one aspect, in step S2, the method comprises the following steps: Step S21: inputting the training data contained in the data set into the initial neural network model; Step S22: Perform forward propagation: perform weighted sum calculation: , transform the activation function: ,in, For the The linear combination result of neurons, n is the number of training samples, For input data, is the weight, is the bias, To be the first The linear combination result of neurons is converted into a nonlinear combination result. is the activation function; Step S23: Calculate the loss function using mean square error: ,in, is the model prediction value, is the actual value; Step S24: using the back propagation method to minimize the loss function through the gradient descent method; Step S25: loop through steps S22 to S24 until optimal parameters are generated that minimize the loss function.

[0008] In one aspect, in step S24, the method comprises the following steps: Step S241: perform gradient calculation: ,in, The loss function is weighted The partial derivative of is the loss function The partial derivative of is the derivative of the activation function, is the partial derivative of the linear combination result with respect to the weight; Step S242: Update weights: in, is the learning rate, which determines the size of each adjustment step. is the updated weight; Step S243: Update the deviation: , is the updated bias.

[0009] In one aspect, in step S2, the method further includes the following steps: Step S26: input the test data of the data set into the trained neural network model; Step S27: Adjust the learning rate, use L2 regularization, and evaluate the regularization effect on the test data; Step S28: Calculate the average error to evaluate the generalization ability of the trained neural network model; Step S29: loop through steps S27 and S28 until a minimum mean square error is generated, and then the neural network model is set as a control prediction model.

[0010] In one aspect, in step S2, the method further comprises the following steps: Step S201: Perform multiple sampling with replacement from the original training data to generate multiple different sub-training data; Step S202: training an independent neural network model on each sub-training data, and then setting each trained neural network model as an independent control prediction model; Step S203: When making a prediction, the outputs of all control prediction models are averaged or majority voted as the final prediction result.

[0011] In one aspect, in step S2, the method further comprises the following steps: Step S204: Initialize the sample weights of the training data to be uniformly distributed; Step S205: In each round of iteration, the neural network model is trained according to the current weight distribution, and the error rate of the model is calculated: ,in, is the error rate of the t-th learner, is the true label of the i-th sample, is the indicator function, when the prediction is wrong ,otherwise , is the weight of the i-th sample in the t-th iteration, is the prediction of the t-th learner for the ith sample.

[0012] Step S206: Calculate the weight of the learner according to the error rate: ,in, is the weight of the t-th learner; Step S207: Update the sample weights so that in the next iteration, the wrongly predicted samples have a larger weight: , normalize all weights; Step S208: Repeat steps S205 to S207 until a predetermined number of iterations or error rate standard is reached; Step S209: weight and calculate all learners to generate the final prediction result.

[0013] In one aspect, in step S4, the method further comprises the following steps: Step S41: real-time collection of multi-dimensional data of tea seed raw materials and input into the control prediction model to generate tea seed prediction data, wherein the tea seed prediction data at least includes: optimal enzyme dosage, pH value, temperature value, crushing parameters, oil pressing parameters and filtering and refining parameters; Step S42: calculating the error between the tea seed real-time condition and the tea seed prediction data according to the tea seed prediction data; Step S43: using a PID control method to dynamically adjust the reaction conditions of the tea seed extraction process.

[0014] In one aspect, in step S43, the method comprises the following steps: Step S431: Use the PID formula to calculate the control strategy in real time: in, is a control signal used to adjust the reaction conditions. is the error between the real-time condition of tea seeds and the predicted data of tea seeds; The proportional, integral and differential gains of PID control are used to control the current error, cumulative error and error change rate; Step S432: adjusting the reaction conditions of the tea seed extraction process in real time according to the control strategy.

[0015] The present application also provides an adaptive control system for improving tea oil extraction efficiency by executing the adaptive control method, comprising: A data processing module is used to perform data preprocessing based on the multi-dimensional data of tea seed raw materials collected historically, and then generate a data set including training data and test data; The model training module is used to continuously train the data set using a neural network to generate a modified control prediction model, wherein the training process includes: forward propagating the training data of the data set and transforming it through weighted summation and activation function, calculating the loss function, performing back propagation, and updating parameters through the gradient descent method; A model prediction module is used to deploy the control prediction model, and then input the collected real-time tea seed raw material multi-dimensional data into the control prediction model to generate tea oil extraction prediction data; The strategy execution module is used to generate an operation strategy for a matching tea oil extraction device based on tea oil extraction prediction data, and then dynamically control the operation parameters of the matching tea oil extraction device according to the operation strategy and in a matching time interval.

[0016] The above technical solution of the present application has the following advantages compared with the prior art: 1. Based on the multidimensional data of tea seed raw materials collected historically, this application constructs a neural network model to predict and control the optimal parameters for tea oil extraction, and then generates the optimal extraction conditions through real-time collected data and the trained control prediction model, which can ensure that the extraction process of each batch of tea seeds is carried out under the optimal conditions, reducing the fluctuation of extraction efficiency caused by fluctuations in raw material properties. In addition, through various historical tea seed raw material multidimensional data, various tea seed types can be identified and processed in advance, adapting to changes in raw materials and environment, and ensuring the stability and consistency of tea oil extraction.

[0017] 2. Based on the prediction model input based on the real-time collected data, the generated tea oil extraction equipment operation strategy can adjust the equipment's operating parameters in real time according to the characteristics of tea seeds, so that the equipment is always in the best working state, maximize the extraction efficiency, and improve the resource utilization of the production process.

[0018] 3. Real-time monitoring of the reaction conditions of tea seeds during the extraction process and feedback of error signals to the controller. The PID control strategy is used to adjust the reaction conditions in real time, which can balance the contradiction between rapid response and stability. This allows for rapid adjustments in response to environmental changes while maintaining high control accuracy to avoid overshoot or oscillation. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on the provided drawings without paying creative work.

[0020] Figure 1 It is a flow chart of an adaptive control method for improving tea oil extraction efficiency provided by an embodiment of the present invention.

[0021] Figure 2 It is a flow chart of a data processing method provided by an embodiment of the present invention.

[0022] Figure 3 It is a flow chart of the model training and evaluation method provided by an embodiment of the present invention.

[0023] Figure 4 It is a flow chart of the parallel ensemble learning method provided by an embodiment of the present invention.

[0024] Figure 5 It is a flow chart of the serial ensemble learning method provided by an embodiment of the present invention.

[0025] Figure 6 It is a flow chart of the enzymatic extraction method provided by an embodiment of the present invention.

[0026] Figure 7 It is a flow chart of a control strategy calculation method provided by an embodiment of the present invention.

[0027] Figure 8 It is a schematic diagram of module connections of an adaptive control system for improving tea oil extraction efficiency provided by an embodiment of the present invention.

[0028] Description of the Figures in the Specification: 101. Data processing module, 102. Model training module, 103. Model prediction module, 104. Strategy execution module. DETAILED DESCRIPTION

[0029] Embodiments of the present invention are described in detail below, examples of which are shown in the accompanying drawings, wherein the same or similar reference numerals throughout represent the same or similar elements or elements having the same or similar functions. The embodiments described below with reference to the accompanying drawings are exemplary and are intended to be used to explain the present invention, and should not be construed as limiting the present invention.

[0030] refer to Figure 1 As shown, in some embodiments, a self-adaptive control method for improving tea oil extraction efficiency is provided, the method comprising the following steps: Step S1: Perform data preprocessing based on the multidimensional data of tea seed raw materials collected historically, and then generate a data set including training data and test data.

[0031] Specifically, refer to Figure 2 As shown, in step S1, the method includes the following steps: Step S11: data cleaning is performed based on the historically collected tea seed characteristic data, extraction process parameters, extraction result data and environmental data, wherein the data cleaning includes: outlier and missing value processing.

[0032] The tea seed characteristic data at least includes the moisture content, oil content, and particle size of the tea seeds; the extraction process parameters at least include temperature, pressure, extraction time, and rotation speed; the extraction result data at least includes oil yield, oil purity, and quality indicators; the environmental data at least includes external conditions such as ambient temperature and humidity. The above historical data covers several batches of data in the past. The specific range of historical data used for training is set and adjusted by the operator according to actual needs to help the model learn the impact of different variables on extraction efficiency.

[0033] Specifically, the Z-score statistical method is usually used to remove outliers, and then the outliers are removed; the method for handling missing values ​​usually uses the mean, median, mode to fill the missing values ​​or directly deletes the sample data corresponding to the missing values.

[0034] Step S12: Standardize the cleaned data to generate multi-dimensional data of tea seed raw materials.

[0035] Among them, after outliers are eliminated and missing values ​​are filled or eliminated, the above-mentioned tea seed characteristic data, extraction process parameters, extraction result data and environmental data are subjected to standard deviation standardization. Through standardization, the features of different dimensions and ranges are unified to the same standard to eliminate the dimensional influence between features and improve the model training effect. The specific process is as follows: Calculate the mean and standard deviation for each feature; Standardize each feature to generate standardized data features; The standardized data features are aggregated into a multidimensional data structure to generate multidimensional data of tea seed raw materials.

[0036] Step S13: dividing the multidimensional data of tea seed raw materials into training data and test data, and then generating a data set including the training data and the test data.

[0037] Among them, after the multidimensional data of tea seed raw materials are standardized, the multidimensional data of tea seed raw materials are divided into training data and test data, which are used for model training and verification respectively. The usual division ratio is 80% training data and 20% test data.

[0038] Step S2: Using a neural network to continuously train the data set to generate a revised control prediction model, the training process includes: forward propagating the training data of the data set and transforming it through weighted summation and activation function, calculating the loss function, and using back propagation to update parameters through the gradient descent method.

[0039] Among them, the neural network model used in this application includes an input layer, a hidden layer and an output layer. The input layer corresponds to the number of features of the input data. For example, assuming that the input features include moisture content, temperature, pressure, rotation speed, etc., the number of neurons in the input layer is consistent with the number of these variables; the hidden layer is used to capture the nonlinear relationship between input and output. The number of hidden layers and the number of nodes in each layer are set by the operator according to actual needs. For example, the usual design method is 2 to 3 hidden layers, with 20-100 neurons in each layer; the output layer corresponds to the number of target variables to be predicted. For example, if it is necessary to predict the extraction efficiency, optimal extraction time and energy consumption, the output layer contains 3 nodes. The activation function of the neural network model used in this application adopts the ReLU (Rectified Linear Unit) function, which can effectively alleviate the gradient vanishing problem when processing large-scale data.

[0040] Specifically, refer to Figure 3 As shown, in step S2, the method further includes the following steps: Step S21: input the training data contained in the data set into the initial neural network model.

[0041] Step S22: Perform forward propagation: perform weighted sum calculation on each neuron: , use the activation function to convert the linear combination result into a nonlinear output: ,in, For the The linear combination result of neurons, n is the number of training samples, For input data, is the weight, is the bias, To be the first The linear combination result of neurons is converted into a nonlinear combination result. is the activation function.

[0042] Step S23: Calculate the loss function using mean square error: ,in, is the loss function value, is the model prediction value, is the actual value.

[0043] Step S24: Use the back-propagation method to minimize the loss function through the gradient descent method.

[0044] Step S25: loop through steps S22 to S24 until optimal parameters are generated that minimize the loss function.

[0045] Specifically, in step S24, the method includes the following steps: Step S241: perform gradient calculation: ,in, The loss function is weighted The partial derivative of is the loss function The partial derivative of is the derivative of the activation function, is the partial derivative of the linear combination result with respect to the weight.

[0046] Step S242: Update weights: in, is the learning rate, which determines the size of each adjustment step. is the updated weight.

[0047] Step S243: Update the deviation: , is the updated bias.

[0048] Step S26: Input the test data of the data set into the trained neural network model to evaluate the generalization ability of the model.

[0049] Step S27: According to the performance of the model on the test set, dynamically adjust the learning rate, add the L2 regularization term to the loss function to prevent overfitting, and evaluate the regularization effect on the test data. The specific process is as follows: in, is the loss function after regularization, is the regularization coefficient, which is used to control the regularization strength. is the number of weights in the model.

[0050] Step S28: Calculate the average error to evaluate the generalization ability of the trained neural network model. The specific calculation process is: in, is the mean square error of the test set, is the number of test data samples.

[0051] Step S29: loop through steps S27 and S28 until a minimum mean square error is generated, and then the neural network model is set as a control prediction model.

[0052] Step S3: deploy the control prediction model, and then input the collected real-time tea seed raw material multi-dimensional data into the control prediction model to generate tea oil extraction prediction data.

[0053] Among them, the generated tea oil extraction prediction data includes but is not limited to tea oil extraction efficiency, optimal extraction time, equipment energy consumption, pH value, extraction temperature, enzyme dosage (such as cellulase, pectinase), etc.

[0054] Step S4: Generate an operation strategy for a matching tea oil extraction device based on the tea oil extraction prediction data, and then dynamically control the operation parameters of the matching tea oil extraction device according to the operation strategy and in a matching time interval.

[0055] Among them, according to the tea oil extraction prediction data, a matching operation strategy is formulated. For example, for example: define target parameters: determine the equipment operation parameters that need to be controlled, such as extraction temperature, extraction pressure, extraction time, and rotation speed; formulate strategy rules: formulate the equipment operation strategy based on the prediction data and preset goals. Strategies can be classified according to extraction efficiency, such as: if the oil yield is higher than the set threshold, reduce energy consumption; if the oil yield is lower than the set threshold, increase the extraction temperature or time. Thus, the operation parameters are calculated. For example: assuming that the set target oil yield is , the following formula can be used to calculate the operating parameters: in, is the adjusted extraction temperature, is the adjusted extraction pressure, is the adjusted extraction time, is the adjusted speed, is the reference parameter of the equipment under standard conditions, is the adjustment coefficient corresponding to the difference in oil yield, indicating the impact of each unit change in oil yield on the parameter.

[0056] Then, the operating parameters of the tea oil extraction equipment are adjusted in real time based on the operating parameters calculated based on the facts.

[0057] In some embodiments, reference Figure 4 As shown, in step S2, the method further includes the following steps: Step S201: Perform multiple sampling with replacement from the original training data to generate multiple different sub-training data.

[0058] A plurality of sub-training data are randomly extracted from the original training data with replacement, and the size of each sub-training data is the same as the original training data.

[0059] Step S202: Train an independent neural network model on each sub-training data, and then set each trained neural network model as an independent control prediction model.

[0060] Among them, an independent neural network model is trained on each sub-training data using the above method.

[0061] Step S203: When making a prediction, the outputs of all control prediction models are averaged or majority voted as the final prediction result.

[0062] For example, the formula for taking the average value is as follows: ,in, is the number of independent models, is the predicted value of the i-th model for the input sample, The final ensemble prediction results.

[0063] Through the above-mentioned parallel ensemble learning method, multiple sub-sample sets are randomly selected from the original data set to train multiple independent models. Finally, the prediction results of these models are averaged or majority voted to reduce the variance of the model, enhance the generalization ability of the model, reduce the risk of overfitting of a single model, and improve the stability of the model.

[0064] In some embodiments, reference Figure 5 As shown, in step S2, the method further includes the following steps: Step S204: Initialize the sample weights of the training data to be uniformly distributed, that is, initialize each training data to have the same weight.

[0065] Step S205: In each round of iteration, the neural network model is trained according to the current weight distribution, and the error rate of the model is calculated: ,in, is the error rate of the t-th learner, is the true label of the i-th sample, is the indicator function, when the prediction is wrong ,otherwise , is the weight of the i-th sample in the t-th iteration, is the prediction of the t-th learner for the i-th sample.

[0066] Step S206: Calculate the weight of the learner according to the error rate: ,in, is the weight of the t-th learner, The smaller the model weight The larger it is, the better the model performance; Step S207: Update the sample weights so that in the next iteration, the wrongly predicted samples have a larger weight: , normalize all weights:

[0067] Step S208: Repeat steps S205 to S207 until a predetermined number of iterations or error rate standard is reached; Step S209: All learners are weighted and calculated to generate the final prediction result: ,in, is the final prediction result, and T is the total number of iterative training rounds.

[0068] Through the above-mentioned serial ensemble learning method, multiple learners are gradually constructed, and each new model focuses on improving the erroneous predictions of the previous model, so that each round of new models can significantly reduce the overall error and improve the prediction accuracy.

[0069] In some embodiments, reference Figure 6 As shown, in step S4, the method further includes the following steps: Step S41: real-time collection of multi-dimensional data of tea seed raw materials and input into the control prediction model to generate tea seed prediction data, wherein the tea seed prediction data at least includes: optimal enzyme dosage, pH value, temperature value, crushing parameters, oil extraction parameters and filtering and refining parameters.

[0070] Among them, the enzymatic method can be used to optimize the extraction efficiency of tea oil. By using specific enzymes (such as cellulase and pectinase) to destroy the cell walls of tea seeds and promote the release of oil, this method can improve the extraction rate of oil, reduce the dependence on high temperature, and retain more active ingredients; the specific steps of the enzymatic method are shown in the following examples: Wash, dry and crush the tea seeds to ensure that the surface of the tea seeds is clean and the particles are uniform, which is conducive to the subsequent enzymatic hydrolysis reaction; Pre-soak the crushed tea seeds with an appropriate amount of aqueous solution to soften the cell wall and reduce the reaction resistance during subsequent enzymatic hydrolysis. At the same time, a predetermined amount of pH regulator can be added to optimize the subsequent reaction environment. Choose enzymes suitable for destroying cell walls, such as cellulase, pectinase, or a combination of enzymes that are actually needed.

[0071] Specifically, the prediction model generates prediction data of the optimal reaction conditions based on the input data, including: optimal enzyme dosage, optimal pH value, optimal reaction temperature, crusher speed and time, oil press pressure and speed, and specific parameters of filtration and refining.

[0072] Step S42: Calculate the error between the tea seed real-time condition and the tea seed prediction data according to the tea seed prediction data.

[0073] Among them, the parameters monitored in real time during the tea seed extraction process are compared with the optimal parameters generated by the prediction model to calculate the error value; for example, the current actual temperature, pH, enzyme amount, etc. may deviate from the optimal values, and these deviations constitute error signals.

[0074] Step S43: using a PID control method to dynamically adjust the reaction conditions of the tea seed extraction process.

[0075] Specifically, refer to Figure 7 As shown, in step S43, the method includes the following steps: Step S431: Use the PID control algorithm to adjust the reaction conditions to minimize the error between the predicted value and the actual value. The specific formula is: in, is a control signal, which indicates the adjustment amount of the reaction conditions in the tea seed extraction process, such as the increase or decrease of temperature, the adjustment of pH value, the increase or decrease of enzyme amount, etc.; is a real-time error signal, which refers to the deviation between the actual conditions measured in the current tea seed extraction process and the predicted data. Its calculation formula is: , are the optimal condition parameters of the tea seed extraction process generated by the control prediction model, It is the actual condition parameters of the tea seed extraction process measured in real time; It is the proportional gain coefficient, which is used to adjust the influence of the current error value on the control signal. Its function is to directly control the size of the current error so that the system can respond quickly when a deviation occurs. It is the integral gain coefficient, which reflects the past error accumulation and is used to eliminate the long-term steady-state error of the system. The integral term accumulates the past errors so that the system can continuously approach the set value even if the error is very small. It is the differential gain coefficient, which reflects the error change rate and is used to predict the future error development trend and adjust the control amount in advance to prevent the error from expanding rapidly. The differential term can slow down the response speed and avoid overshoot of the system.

[0076] Step S432: adjusting the reaction conditions of the tea seed extraction process in real time according to the control strategy.

[0077] Specifically, the control signal calculated , real-time adjustment of various reaction conditions in the tea seed extraction process, such as: adjusting the reaction temperature to make it tend to the optimal temperature; adjusting the pH value to keep it in the optimal range; automatically adjusting the enzyme dosage to ensure reaction efficiency; optimizing the parameters of the crusher, oil press and filtering and refining equipment to achieve the best extraction effect.

[0078] The tea seed cell walls are destroyed by enzymatic method. The key parameters of the enzymatic hydrolysis process (such as enzyme activity, pH, temperature, reaction time, etc.) are monitored in real time through an adaptive control system, and the process parameters are automatically adjusted to ensure that the enzymatic reaction is always in the best state to improve the oil extraction rate.

[0079] In some embodiments, reference Figure 8 As shown, the present application also relates to an adaptive control system for improving tea oil extraction efficiency by executing the adaptive control method, comprising: A data processing module is used to perform data preprocessing based on the multi-dimensional data of tea seed raw materials collected historically, and then generate a data set including training data and test data; The model training module is used to continuously train the data set using a neural network to generate a modified control prediction model, wherein the training process includes: forward propagating the training data of the data set and transforming it through weighted summation and activation function, calculating the loss function, performing back propagation, and updating parameters through the gradient descent method; A model prediction module is used to deploy the control prediction model, and then input the collected real-time tea seed raw material multi-dimensional data into the control prediction model to generate tea oil extraction prediction data; The strategy execution module is used to generate an operation strategy for a matching tea oil extraction device based on tea oil extraction prediction data, and then dynamically control the operation parameters of the matching tea oil extraction device according to the operation strategy and in a matching time interval.

[0080] In some embodiments, the present application also relates to a computer medium having a computer program stored thereon, and the computer program is executed by a processor to implement the adaptive control method for improving tea oil extraction efficiency according to the above-mentioned method.

[0081] In some embodiments, the present application also relates to a computer, comprising the above-mentioned computer medium.

[0082] In the description of this specification, the description with reference to the terms "one embodiment", "some embodiments", "example", "specific example", or "some examples" etc. means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present invention. In this specification, the schematic representations of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described may be combined in any one or more embodiments or examples in a suitable manner. In addition, those skilled in the art may combine and combine the different embodiments or examples described in this specification and the features of the different embodiments or examples, without contradiction.

[0083] Although the embodiments of the present invention have been shown and described above, it is to be understood that the above embodiments are exemplary and are not to be construed as limitations of the present invention. A person skilled in the art may change, modify, replace and vary the above embodiments within the scope of the present invention.

Claims

1. An adaptive control method for improving tea oil extraction efficiency, characterized in that: The following steps are involved: Step S1: performing data preprocessing based on the multidimensional data of tea seed raw materials collected historically, thereby generating a data set including training data and test data; Step S2: using a neural network to continuously train the data set to generate a revised control prediction model, the training process includes: forward propagating the training data of the data set and transforming it through weighted summation and activation function, calculating the loss function, and using back propagation to update parameters through gradient descent method; Step S3: deploying the control prediction model, and then inputting the collected real-time tea seed raw material multi-dimensional data into the control prediction model to generate tea oil extraction prediction data; Step S4: Generate an operation strategy for a matching tea oil extraction device based on the tea oil extraction prediction data, and then dynamically control the operation parameters of the matching tea oil extraction device according to the operation strategy and in a matching time interval.

2. The adaptive control method for improving tea oil extraction efficiency according to claim 1, characterized in that: In step S1, the method comprises the following steps: Step S11: performing data cleaning based on historically collected tea seed characteristic data, extraction process parameters, extraction result data and environmental data, wherein the data cleaning includes: outlier and missing value processing; Step S12: standardizing the cleaned data to generate multi-dimensional data of tea seed raw materials; Step S13: dividing the multidimensional data of tea seed raw materials into training data and test data, and then generating a data set including the training data and the test data.

3. The adaptive control method for improving tea oil extraction efficiency according to claim 1, characterized in that: In step S2, the method comprises the following steps: Step S21: inputting the training data contained in the data set into the initial neural network model; Step S22: Perform forward propagation: perform weighted sum calculation: ， Transform the activation function: ， in, For the The linear combination result of neurons, n is the number of training samples, For input data, is the weight, is the bias, To be the first The linear combination result of neurons is converted into a nonlinear combination result. is the activation function; Step S23: Calculate the loss function using mean square error: ， in, is the model prediction value, is the actual value; Step S24: using the back propagation method to minimize the loss function through the gradient descent method; Step S25: loop through steps S22 to S24 until optimal parameters are generated that minimize the loss function.

4. The adaptive control method for improving tea oil extraction efficiency according to claim 3, characterized in that: In step S24, the method comprises the following steps: Step S241: perform gradient calculation: ， in, The loss function is weighted The partial derivative of is the loss function The partial derivative of is the derivative of the activation function, is the partial derivative of the linear combination result with respect to the weight; Step S242: Update weights: in, is the learning rate, which determines the size of each adjustment step. is the updated weight; Step S243: Update the deviation: ， is the updated bias.

5. The adaptive control method for improving tea oil extraction efficiency according to claim 4, characterized in that: In step S2, the method further includes the following steps: Step S26: input the test data of the data set into the trained neural network model; Step S27: Adjust the learning rate, use L2 regularization, and evaluate the regularization effect on the test data; Step S28: Calculate the average error to evaluate the generalization ability of the trained neural network model; Step S29: loop through steps S27 and S28 until a minimum mean square error is generated, and then the neural network model is set as a control prediction model.

6. The adaptive control method for improving tea oil extraction efficiency according to claim 1, characterized in that: In step S2, the method further comprises the following steps: Step S201: Perform multiple sampling with replacement from the original training data to generate multiple different sub-training data; Step S202: training an independent neural network model on each sub-training data, and then setting each trained neural network model as an independent control prediction model; Step S203: When making a prediction, the outputs of all control prediction models are averaged or majority voted as the final prediction result.

7. The adaptive control method for improving tea oil extraction efficiency according to claim 1, characterized in that: In step S2, the method further comprises the following steps: Step S204: Initialize the sample weights of the training data to be uniformly distributed; Step S205: In each round of iteration, the neural network model is trained according to the current weight distribution, and the error rate of the model is calculated: ， in, is the error rate of the t-th learner, is the true label of the i-th sample, is the indicator function, when the prediction is wrong ,otherwise , is the weight of the i-th sample in the t-th iteration, is the prediction of the t-th learner for the i-th sample; Step S206: Calculate the weight of the learner according to the error rate: ， in, is the weight of the t-th learner; Step S207: Update the sample weights so that in the next iteration, the wrongly predicted samples have a larger weight: ， Normalize all weights; Step S208: Repeat steps S205 to S207 until a predetermined number of iterations or error rate standard is reached; Step S209: weight and calculate all learners to generate the final prediction result.

8. The adaptive control method for improving tea oil extraction efficiency according to claim 1, characterized in that: In step S4, the method further comprises the following steps: Step S41: real-time collection of multi-dimensional data of tea seed raw materials and input into the control prediction model to generate tea seed prediction data, wherein the tea seed prediction data at least includes: optimal enzyme dosage, pH value, temperature value, crushing parameters, oil pressing parameters and filtering and refining parameters; Step S42: calculating the error between the tea seed real-time condition and the tea seed prediction data according to the tea seed prediction data; Step S43: using a PID control method to dynamically adjust the reaction conditions of the tea seed extraction process.

9. The adaptive control method for improving tea oil extraction efficiency according to claim 8, characterized in that: In step S43, the method comprises the following steps: Step S431: Use the PID formula to calculate the control strategy in real time: in, is a control signal used to adjust the reaction conditions. is the error between the real-time condition of tea seeds and the predicted data of tea seeds; The proportional, integral and differential gains of PID control are used to control the current error, cumulative error and error change rate; Step S432: adjusting the reaction conditions of the tea seed extraction process in real time according to the control strategy.

10. An adaptive control system for improving tea oil extraction efficiency by executing the adaptive control method according to any one of claims 1 to 9, characterized in that: include: A data processing module is used to perform data preprocessing based on the multi-dimensional data of tea seed raw materials collected historically, and then generate a data set including training data and test data; The model training module is used to continuously train the data set using a neural network to generate a modified control prediction model, wherein the training process includes: forward propagating the training data of the data set and transforming it through weighted summation and activation function, calculating the loss function, performing back propagation, and updating parameters through the gradient descent method; A model prediction module is used to deploy the control prediction model, and then input the collected real-time tea seed raw material multi-dimensional data into the control prediction model to generate tea oil extraction prediction data; The strategy execution module is used to generate an operation strategy for a matching tea oil extraction device based on tea oil extraction prediction data, and then dynamically control the operation parameters of the matching tea oil extraction device according to the operation strategy and in a matching time interval.

Images