Intelligent optimization method and system for seabuckthorn full-ingredient extraction process

Abstract

The present invention relates to the technical field of seabuckthorn full component extraction, and discloses a seabuckthorn full component extraction process intelligent optimization method and system, including obtaining a feature set of the seabuckthorn full component extraction process, constructing new features, preprocessing the feature set and the new features, encoding the seabuckthorn variety, origin, similarity between varieties and correlation between varieties and origins, constructing a feature matrix, predicting the extraction product type and extraction rate based on the feature matrix and a seabuckthorn full component extraction prediction model, and dynamically optimizing the seabuckthorn full component extraction process based on the prediction results. The present invention can realize the automation and intelligent optimization of the seabuckthorn extraction process, and improve production efficiency and quality stability.

Description

Technical Field

[0001] The present invention relates to the technical field of seabuckthorn full component extraction, and more specifically, to an intelligent optimization method and system for seabuckthorn full component extraction process. Background Art

[0002] At present, seabuckthorn extraction mainly adopts supercritical CO2 extraction, solvent extraction and water extraction, which have problems such as low extraction efficiency, high energy consumption and solvent residue. In addition, the properties of seabuckthorn raw materials from different varieties and origins vary greatly, and fixed extraction processes are difficult to adapt to variable raw materials, resulting in unstable extraction rate and product quality. Therefore, it is urgent to develop an intelligent seabuckthorn full-ingredient extraction optimization method that can dynamically adjust process parameters according to raw material characteristics and target products to achieve automatic optimization control of the extraction process. Summary of the invention

[0003] In order to overcome the above-mentioned defects of the prior art, the present invention provides a method and system for intelligent optimization of the seabuckthorn full-ingredient extraction process. The seabuckthorn full-ingredient extraction intelligent optimization method of the present invention solves the problems of incomplete original data and poor quality by comprehensively collecting key features of the extraction process, constructing new features such as pressure-temperature ratio, solvent polarity index, etc., and performing data preprocessing. The introduction of heterogeneous graphs and graph neural networks to model the correlation between seabuckthorn varieties and origins overcomes the shortcomings of existing methods that ignore the differences in raw materials. A seabuckthorn full-ingredient extraction prediction model is constructed to achieve accurate prediction of the extraction product type and extraction rate, and the extraction process parameters are optimized and adjusted in real time based on deviation feedback, which greatly improves production efficiency and quality stability compared to traditional methods, and has significant technical and economic benefits.

[0004] To achieve the above object, the present invention provides the following technical solutions:

[0005] An intelligent optimization method for extracting full components of seabuckthorn, comprising:

[0006] S10: Obtaining a feature set of the seabuckthorn full component extraction process, wherein the feature set includes continuous features and discrete features;

[0007] S20: constructing new features of the seabuckthorn full component extraction process, the new features including pressure-temperature ratio, solvent polarity index and volume-surface area ratio;

[0008] S30: preprocessing the feature set and new features of the seabuckthorn full component extraction process;

[0009] S40: coding of sea buckthorn varieties, origins, similarities between varieties, and associations between varieties and origins;

[0010] S50: construct a feature matrix, and predict the extraction product type and extraction rate based on the feature matrix and the seabuckthorn full component extraction prediction model;

[0011] S60: Dynamically optimize the seabuckthorn full ingredient extraction process based on predicted extraction product types and extraction rates.

[0012] Furthermore, the process of preprocessing the feature set and new features of the seabuckthorn full component extraction process includes: outlier processing, standardization processing and category coding; the outlier processing method includes a weighted outlier processing method; the standardization processing method includes an adaptive standardization processing method;

[0013] The weighted outlier processing method includes: calculating the weighted Euclidean distance of samples in the feature vector, calculating the mean and standard deviation of the weighted Euclidean distance of all samples, and judging and processing outliers;

[0014] The method for calculating the weighted Euclidean distance of samples in the feature vector includes:

[0015]

[0016] in, is the weighted Euclidean distance of the i-th sample, x is the feature vector of the continuous feature, n is the number of samples, p is the number of features, is the kth eigenvalue of the ith sample, is the kth eigenvalue of the jth sample, For sample The weight of .

[0017] Furthermore, the method for judging and processing outliers includes:

[0018]

[0019] in, is the mean of the sample weighted Euclidean distance, is the standard deviation of the sample weighted Euclidean distance, is the weighted Euclidean distance of the ith sample, z is the z-score value; set is the abnormal threshold;

[0020] like ,but For outliers, after detecting them, they are removed or replaced with the average value of other samples;

[0021] The adaptive normalization processing method comprises:

[0022]

[0023]

[0024] in, is the adaptive parameter, is the kth eigenvalue of the ith sample, is the standardized value of the kth feature of the i-th sample, is the mean of the kth feature, is the standard deviation of the kth feature.

[0025] Furthermore, the method for encoding seabuckthorn varieties, origins, similarities between varieties, and associations between varieties and origins includes: constructing a seabuckthorn variety-origin heterogeneous graph, using a graph neural network algorithm to learn embedded representations of variety and origin nodes, and obtaining a final variety-origin representation;

[0026] The method for constructing a seabuckthorn variety-origin heterogeneous graph comprises: obtaining variety-variety similarity edge information and obtaining variety-origin association edge information;

[0027] The method for obtaining variety-variety similarity edge information includes: obtaining genome information of different seabuckthorn varieties, calculating the genetic distance between different seabuckthorn varieties according to the genome information, and obtaining the similarity between the varieties based on the genetic distance.

[0028] Furthermore, the method for obtaining the variety-origin association edge information includes: counting the planting area proportion of each seabuckthorn variety in different origins, and constructing a A matrix A, where is the number of varieties, q is the number of origins, Indicates The species in The proportion of planting area in each production area;

[0029] The method for calculating the genetic distance between different seabuckthorn varieties comprises:

[0030] Assume that There are 100 varieties of sea buckthorn, and the genome of each variety is represented by a binary vector of length s, denoted as ,in, ; Indicates the variety There is a variation at the rth gene locus, Indicates no variation, variety and varieties The genetic distance between Use the following formula to measure:

[0031]

[0032] in represents the number of intersections between two seabuckthorn varieties at the mutation sites, represents the number of unions of two seabuckthorn varieties at the variant sites;

[0033] The method for obtaining the similarity between varieties based on genetic distance comprises:

[0034]

[0035] in, For varieties and varieties The similarity between is the bandwidth parameter of the Gaussian kernel, is an exponential function.

[0036] Furthermore, the method for dynamically optimizing the process flow of sea buckthorn full-ingredient extraction based on predicted extraction product type and extraction rate includes: comparing the predicted extraction product type and extraction rate with the target extraction product type and extraction rate, calculating the deviation, and obtaining the product type prediction deviation vector and the extraction rate prediction deviation; calculating the control amount of the process parameters based on the product type prediction deviation vector and the extraction rate prediction deviation; performing decoupling calculation on the control amount of the process parameters to obtain the process parameter adjustment amount, applying the calculated process parameter adjustment amount to the actual production process, and dynamically optimizing the extraction process.

[0037] Furthermore, the method of comparing the predicted extracted product type and extraction rate with the target extracted product type and extraction rate, calculating the deviation, and obtaining the product type prediction deviation vector and the extraction rate prediction deviation includes:

[0038]

[0039]

[0040] in, Extract the encoding vector of product type for the target, is the target extraction rate, is the product type prediction deviation vector, is the extraction rate prediction deviation, Encode the vector for the predicted extracted product type, is the predicted extraction rate;

[0041] The method for calculating the control amount of the process parameter based on the product type prediction deviation vector and the extraction rate prediction deviation includes:

[0042]

[0043]

[0044] in, For the current moment, is the control quantity vector of the process parameters at the current moment, The current moment is the comprehensive deviation, For a moment in the past The combined deviation of is the weight matrix of product type prediction bias, Forecast deviation balance factor for product type, is the extraction rate prediction deviation balance factor, is the proportionality coefficient, is the integration coefficient, is the differential coefficient, is the prediction deviation vector of the product type at the current moment, The predicted deviation of the extraction rate at the current moment.

[0045] A seabuckthorn full-ingredient extraction process intelligent optimization system, which is used to implement the above-mentioned seabuckthorn full-ingredient extraction process intelligent optimization method, comprising:

[0046] Process feature set acquisition module: used to comprehensively collect various key features that affect seabuckthorn extraction results, including extraction methods, extraction conditions, pretreatment processes, seabuckthorn raw materials, seabuckthorn varieties and origins, etc., to lay a data foundation for subsequent analysis and avoid missing important information;

[0047] New feature construction module: used to obtain new features in the process parameters of seabuckthorn full-ingredient extraction, including pressure-temperature ratio, solvent polarity index and volume-surface area ratio, enrich the original feature space, provide more dimensional information for the model, and help improve prediction accuracy and generalization ability;

[0048] Data preprocessing module: used to process outliers, standardize and encode the feature set and new features of the seabuckthorn full component extraction process to improve data quality;

[0049] Seabuckthorn ingredient extraction prediction module: used to construct a feature matrix, and predict the extraction product type and extraction rate based on the feature matrix and the seabuckthorn full ingredient extraction prediction model;

[0050] Process parameter dynamic optimization module: It is used to dynamically optimize the seabuckthorn full-ingredient extraction process according to the predicted seabuckthorn extraction product type and extraction rate, so that the extraction process can be automated and intelligent, and the production efficiency and quality stability can be improved.

[0051] An electronic device comprises a memory, a central processing unit and a computer program stored in the memory and executable on the central processing unit. When the central processing unit executes the computer program, the above-mentioned intelligent optimization method for extracting the full components of seabuckthorn is realized.

[0052] A computer-readable storage medium having a computer program stored thereon, wherein the computer program, when executed, implements the above-mentioned intelligent optimization method for extracting the full components of seabuckthorn.

[0053] Compared with the prior art, the present invention has the following beneficial effects:

[0054] The present invention solves the problems of incomplete original data and poor quality by comprehensively collecting key features of the seabuckthorn extraction process, constructing new features, and performing data preprocessing, thus laying a solid data foundation for subsequent analysis.

[0055] The present invention innovatively introduces heterogeneous graphs and graph neural networks to model the similarities and correlations between seabuckthorn varieties and origins, overcoming the shortcomings of existing methods that ignore the differences in raw materials, making full use of raw material attribute information, and improving the pertinence of optimization decisions.

[0056] The present invention constructs a prediction model for the extraction of all seabuckthorn components, which can achieve accurate prediction of the type of extracted products and extraction rate, and provide a reliable basis for dynamic optimization of the extraction process. Based on the feedback of model prediction deviation, the extraction process parameters are adjusted in real time, realizing the automated and intelligent closed-loop optimization control of the extraction process.

[0057] Compared with the traditional fixed process, the present invention can flexibly adjust the process parameters according to the characteristics of the raw materials and the target products, thereby improving the extraction efficiency, reducing energy consumption, reducing solvent residues, and improving the stability of product quality, thus having significant technical and economic benefits. BRIEF DESCRIPTION OF THE DRAWINGS

[0058] Figure 1 A flow chart of an intelligent optimization method for extracting all seabuckthorn ingredients in the present invention;

[0059] Figure 2 It is a schematic diagram of the composition structure of the prediction model for extracting all seabuckthorn ingredients in the present invention;

[0060] Figure 3 This is a functional module diagram of an intelligent optimization system for the seabuckthorn full-ingredient extraction process in the present invention. DETAILED DESCRIPTION

[0061] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0062] Example 1

[0063] See also Figure 1 As shown, this embodiment provides an intelligent optimization method for the extraction process of seabuckthorn full ingredients, comprising:

[0064] S10: Obtaining a feature set of the seabuckthorn full component extraction process, wherein the feature set includes continuous features and discrete features;

[0065] In the process of extracting the full components of seabuckthorn, the process parameters that affect the extraction results mainly include extraction method, extraction conditions, pretreatment process, seabuckthorn raw materials, seabuckthorn varieties and origins, etc.

[0066] Extraction methods include: supercritical CO2 extraction, solvent extraction, water extraction, etc.;

[0067] The extraction conditions include: temperature, pressure, solvent type, solid-liquid ratio, extraction time, ultrasonic intensity and microwave power, etc.;

[0068] Pretreatment processes include: drying, crushing, alcohol extraction, enzymatic hydrolysis, etc.

[0069] Sea buckthorn raw materials include: fresh fruit, pomace, seeds and other different parts.

[0070] The continuous type characteristics include: temperature, pressure, material-liquid ratio, extraction time, ultrasonic intensity and microwave power, etc.;

[0071] The discrete characteristics include: extraction method, solvent type, pretreatment process, seabuckthorn raw material, seabuckthorn variety and origin, etc.

[0072] The supercritical CO2 extraction is more suitable for extracting non-polar substances, such as fat-soluble vitamins, fatty acids, etc., and the extraction rate is relatively high; the solvent extraction can extract polar and non-polar substances, and the extraction rate is high, but the residual solvent is difficult to remove; the water extraction method mainly extracts water-soluble components, such as sugars, phenolic compounds, etc., and the extraction rate is average. Different extraction methods can be used in series, combined or nested.

[0073] It is understandable that during the extraction of all seabuckthorn ingredients, increased temperature can increase the solubility and diffusion coefficient of the substance, thereby increasing the extraction rate, but high temperature may also cause the degradation of heat-sensitive substances; increased pressure is conducive to increasing CO2 density and increasing extraction capacity, but too high pressure will increase energy consumption; polar solvents can extract polar substances, and non-polar solvents are suitable for extracting non-polar substances; increasing the solid-liquid ratio can increase the extraction rate, but too high a ratio will waste solvent; extending the extraction time can increase the extraction rate, but the efficiency gradually decreases; ultrasound and microwaves can promote mass transfer and diffusion of substances and improve extraction efficiency; drying can remove moisture and avoid side reactions such as hydrolysis and enzymolysis; crushing can increase the specific surface area and promote the dissolution of substances; enzymolysis can degrade some high-molecular compounds and increase the dissolution of effective ingredients; fresh fruit contains more water-soluble and heat-sensitive components; pomace is mainly rich in high-molecular compounds such as cellulose and protein; seeds mainly contain fat-soluble components such as oil and vitamin E.

[0074] This step comprehensively collects the characteristics of various key factors that affect the extraction results, including extraction methods, extraction conditions, pretreatment processes, seabuckthorn raw materials, seabuckthorn varieties and origins, etc., to lay a data foundation for subsequent analysis and avoid missing important information. Ultrasonic and microwave assistance are added to the conventional seabuckthorn full-ingredient extraction process to further improve the extraction efficiency.

[0075] S20: constructing new features of the seabuckthorn full component extraction process, the new features including pressure-temperature ratio, solvent polarity index and volume-surface area ratio;

[0076] The pressure-temperature ratio, the solvent polarity index and the volume-surface area ratio are all continuous characteristics;

[0077] In the supercritical CO2 extraction process, temperature and pressure are key factors affecting the extraction effect. According to the ideal gas state equation, it can be seen that there is a certain correlation between pressure and temperature. In order to better reflect this correlation, this embodiment constructs a new feature, pressure-temperature ratio;

[0078] The method for constructing the pressure-temperature ratio includes:

[0079]

[0080] Wherein, PRT is the pressure-temperature ratio, P is the pressure during the extraction of seabuckthorn full components, and T is the temperature.

[0081] The PRT feature reflects the intensity of the effect of pressure on the extraction process of seabuckthorn full components at a given temperature. The larger the PRT value, the higher the pressure and the stronger the CO2 solubility under the same temperature conditions, which is conducive to improving the extraction efficiency. The introduction of the PRT feature can provide the model with more information about the interaction between temperature and pressure.

[0082] In the solvent extraction process, the properties of the solvent, such as polarity and viscosity, have an important influence on the extraction effect. In order to quantify the influence of the solvent properties on the extraction process, this embodiment introduces the solvent polarity index PI as a new feature.

[0083] Exemplarily, three solvents are used: ethanol, n-hexane and water, whose corresponding polarity indexes are 5.2, 0.1 and 10.2, respectively. The polarity index PI is introduced into the characteristic matrix. The polarity index can be obtained using different experimental methods or calculation methods, such as fluorescence method, chromatography or theoretical calculation, etc.:

[0084] Ethanol, PI = 5.2;

[0085] n-Hexane, PI=0.1;

[0086] Water, PI = 10.2;

[0087] The higher the PI, the stronger the polarity of the solvent, which is easier to interact with the polar components in seabuckthorn (such as flavonoids, sugars, etc.), which is beneficial to the extraction of these components. Introducing PI features can help the model understand the mapping relationship between solvent properties and extraction effects.

[0088] The properties of seabuckthorn raw materials such as moisture content and particle size will also affect the extraction efficiency. This embodiment introduces the volume surface area ratio VSAR of the raw materials as a new feature, and its calculation formula is:

[0089]

[0090] Where VSAR is the volume surface area ratio, is the volume of the raw material, SA is the surface area of ​​the raw material;

[0091] VSAR reflects the particle size and specific surface area of ​​the raw material. The smaller the VSAR, the smaller the particle size of the raw material, the larger the specific surface area, and the larger the contact area with the solvent, which is conducive to improving mass transfer efficiency and extraction rate. The introduction of VSAR characteristics can provide the model with information on the association between raw material properties and extraction effects.

[0092] The introduction of new features of the seabuckthorn full-ingredient extraction process in this step enriches the original feature space, provides more dimensional information for the model, and is conducive to improving prediction accuracy and generalization ability.

[0093] S30: preprocessing the feature set and new features of the seabuckthorn full component extraction process;

[0094] The preprocessing process includes: outlier processing, standardization processing and category coding;

[0095] Step S30 in the embodiment of the present application also includes:

[0096] S31: Outlier processing of feature sets and new features of the seabuckthorn full component extraction process;

[0097] The purpose of outlier processing is to detect and process outliers and improve data quality. Traditional outlier processing methods, such as the 3σ principle, treat all outliers equally. However, in actual processes, some outliers may contain valuable information and should not be directly eliminated. Therefore, preferably, this embodiment uses a weighted outlier processing method to process outliers for continuous features (such as temperature, pressure, etc.).

[0098] Furthermore, step S31 of the embodiment of the present application further includes:

[0099] S311: Calculate the weighted Euclidean distance of samples in the feature vector;

[0100] Eigenvectors of continuous features , calculate the samples in each feature vector The weighted Euclidean distance of :

[0101]

[0102] in, is the weighted Euclidean distance of the i-th sample, n is the number of samples, p is the number of features, is the kth eigenvalue of the ith sample, is the kth eigenvalue of the jth sample, For sample The weight can be set according to factors such as sample reliability and importance;

[0103] S312: Calculate the mean and standard deviation of the weighted Euclidean distance of all samples;

[0104] Calculate the mean of the weighted Euclidean distances of all samples and standard deviation The methods include:

[0105]

[0106]

[0107] Where n is the number of samples, is the weighted Euclidean distance of the i-th sample;

[0108] S313: judging outliers based on the mean and standard deviation of the sample weighted Euclidean distance, and processing the outliers;

[0109] Methods for judging and handling outliers include:

[0110] The z-score is a standardized indicator used to measure the difference between a data point and the mean.

[0111] The formula for calculating the z value is:

[0112]

[0113] like ,but is an outlier, among which is the abnormal threshold, usually 2 or 3; after the abnormal value is detected, it will be removed or replaced by the average value of other samples.

[0114] Weighted outlier processing can reduce the negative impact of outliers on subsequent analysis and modeling, and improve the reliability of the results. By weighted distance, we can make full use of prior knowledge and give different degrees of importance to different samples. Its flexible processing method (elimination or correction) can retain useful information to the maximum extent.

[0115] S32: Standardize the feature set and new features of the seabuckthorn full component extraction process;

[0116] The purpose of the standardization process is to eliminate the dimensional influence of continuous features. The traditional z-score standardization method uses the same standardization formula for all features, converting their means and variances into 0 and 1 respectively. This processing method ignores the differences in importance, stability, numerical range, etc. of different features, which may lead to the loss of useful information or the amplification of noise. Therefore, preferably, this embodiment adopts an adaptive standardization processing method, which includes:

[0117]

[0118] in, is an adaptive parameter that assigns different standardized strengths to different features. is the kth eigenvalue of the ith sample, is the standardized value of the kth feature of the i-th sample, is the mean of the kth feature, is the standard deviation of the kth feature;

[0119] The value of can be determined according to the importance and stability of the feature. Preferably, this embodiment calculates The method is:

[0120]

[0121] Right now The value of is proportional to the degree of feature dispersion. Features with high dispersion (i.e. large variance) have is also larger, and the normalization strength is relatively weak, so more original information is retained after normalization.

[0122] The adaptive normalization processing method makes continuous features of different dimensions comparable, which facilitates subsequent analysis and modeling; the adaptive parameters can dynamically adjust the normalization strength according to factors such as the importance and stability of the features, and can also retain more statistical characteristics of the original data and reduce information loss.

[0123] S33: Category coding of feature sets and new features of the seabuckthorn full component extraction process;

[0124] For discrete features such as extraction method, pretreatment process, sea buckthorn raw material, etc., we need to perform category encoding. Preferably, this embodiment uses One-hot encoding, that is, each category is converted into a binary feature;

[0125] For example, taking the extraction method as an example, the three commonly used extraction methods are supercritical CO2 extraction, solvent extraction and water extraction. After One-hot encoding, we get the corresponding three features:

[0126] Supercritical CO2 extraction, ;

[0127] Solvent extraction, ;

[0128] Water extraction, ;

[0129] Each feature indicates whether the sample uses the corresponding extraction method. The advantage of One-hot encoding is that it is simple and intuitive, and does not introduce numerical size relationships. It provides a suitable input format for the neural network and improves the expression ability of the model.

[0130] S40: coding of sea buckthorn varieties, origins, similarities between varieties, and associations between varieties and origins;

[0131] The nutrient content of seabuckthorn fruits of different varieties and origins may vary significantly. Although traditional encoding methods (such as One-hot encoding) can convert discrete variables such as varieties and origins into numerical features, they cannot mine the association information between different varieties and origins. This embodiment adopts a node embedding algorithm based on a graph neural network, taking different varieties and origins as nodes in the graph network, and the associations between them (such as the kinship of varieties, the geographical location similarity of origins, etc.) as edges. The low-dimensional vector representation of each variety and origin is learned through the graph embedding algorithm, which can better characterize the similarities and differences of different varieties and origins in the feature space, providing richer information for the model.

[0132] Step S40 in the embodiment of the present application also includes:

[0133] S41: Constructing a sea buckthorn variety-origin heterogeneity map;

[0134] The sea buckthorn variety-origin heterogeneous graph G=(V, E), wherein V is a node set including two types of nodes: variety nodes and origin nodes; and E is an edge set including two types of edges: variety-variety similarity edges and variety-origin association edges.

[0135] Furthermore, step S41 of the embodiment of the present application further includes:

[0136] S411: Obtaining variety-variety similarity edge information, including: obtaining genome information of different seabuckthorn varieties, calculating genetic distances between different seabuckthorn varieties based on the genome information, and obtaining similarities between varieties based on the genetic distances;

[0137] The method for calculating the genetic distance between different seabuckthorn varieties based on genome information includes:

[0138] Assume that There are 100 varieties of sea buckthorn, and the genome of each variety is represented by a binary vector of length s, denoted as ,in, ; Indicates the variety There is a variation at the rth gene locus, Indicates no variation, variety and varieties The genetic distance between Use the following formula to measure:

[0139]

[0140] in represents the number of intersections between two seabuckthorn varieties at the mutation sites, It represents the union number of two seabuckthorn varieties at the mutation sites.

[0141] The significance of this step is that by calculating the genetic distance between varieties, we can quantitatively characterize the differences between different varieties at the genetic level, providing a basis for subsequent similarity calculations. This distance measurement based on genomic information can better reflect the intrinsic differences between varieties than simple phenotypic characteristics (such as fruit size, seed color, etc.).

[0142] The method for obtaining the similarity between varieties based on genetic distance comprises:

[0143]

[0144] in, For varieties and varieties The similarity between is an exponential function, is the bandwidth parameter of the Gaussian kernel, which controls the decay speed of the similarity;

[0145] The purpose of this step is to map the genetic distance value to the similarity value of [0,1]. The smaller the distance, the higher the similarity between varieties. The form of Gaussian kernel can flexibly control the decay rate of similarity, and the bandwidth parameter The larger it is, the slower the similarity decays, and the less obvious the similarity distinction between varieties is; The smaller it is, the faster the similarity decays, and the more obvious the similarity distinction between varieties is. , while retaining the differences between varieties, it does not over-amplify the impact of noise.

[0146] S412: Obtaining variety-origin association information;

[0147] The method for obtaining the variety-origin association edge information includes: counting the planting area proportion of each seabuckthorn variety in different origins, and constructing a A matrix A, where is the number of varieties, q is the number of origins, Indicates The species in The proportion of planted area in each production area.

[0148] The purpose of this step is to quantify the strength of the association between varieties and origins. The higher the proportion of planted area, the better the adaptability of the variety to the origin, and the greater the weight of the corresponding variety-origin association edge in the heterogeneous graph. This association information based on the proportion of planted area can reflect the preferences of different sea buckthorn varieties for environmental conditions and help understand the intrinsic connection between varieties and origins. In the subsequent graph embedding process, the weight information of the association edge can guide the algorithm to more accurately learn the low-dimensional vector representation of the variety and origin nodes, so that in the embedded space, varieties and origins with strong associations can be closer.

[0149] S42: Use the graph neural network algorithm to learn the embedded representation of the variety and origin nodes to obtain the final variety-origin representation;

[0150] The embedding representation method of the graph neural network algorithm learning the variety and origin nodes includes:

[0151]

[0152] in, For variety nodes of Layer embedding, For variety nodes of Layer embedding, For variety nodes of Layer embedding, For Node The neighbor set, AGG is the aggregation function (such as MEAN, MAX, etc.), CONCAT is the concatenation function, is the learnable transformation matrix and ReLU is the activation function.

[0153] The key here is how to aggregate the information of heterogeneous neighbors. Different neighbors can be distinguished based on the type of edge. For a variety node, its neighbors include similar variety nodes and related origin nodes; for an origin node, its neighbors only include related variety nodes. When aggregating, the embeddings of different types of neighbors can be calculated separately, and then spliced:

[0154]

[0155] in, For Node Neighbors of species, For Node Neighbors of origin.

[0156] Through this heterogeneous neighbor aggregation approach, both the similarity between varieties and the association between variety and origin can be encoded in the embedded representation.

[0157] The final variety-origin is expressed as:

[0158]

[0159]

[0160] is the embedding representation of the variety node, is the embedding representation of the origin node, and With original breed characteristics and original origin characteristics Splice to get the final variety representation And origin indication ;

[0161] In this way, the originally discrete varieties and origins are mapped into a low-dimensional continuous space, and the similarities between varieties and the correlations between varieties and origins are implicitly encoded into the node embedding, which can better characterize the characteristics of varieties and origins.

[0162] This method uses heterogeneous graph embedding technology to characterize the characteristics of seabuckthorn varieties and origins from multiple dimensions such as genome and planting distribution. Compared with simple phenotypic characteristics or artificially designed features, this data-driven representation learning method can automatically extract more essential and comprehensive features and reduce the introduction of human bias.

[0163] S50: construct a feature matrix, and predict the extraction product type and extraction rate based on the feature matrix and the seabuckthorn full component extraction prediction model;

[0164] The purpose of constructing the feature matrix is ​​to integrate all features into a standardized numerical matrix as the input layer of the neural network.

[0165] The information in the feature matrix includes the preprocessed continuous features and the encoded discrete features; the discrete features include the extraction method, solvent type and other features encoded by one-hot encoding, and the sea buckthorn variety and origin features encoded by a node embedding algorithm based on a graph neural network;

[0166] The method for constructing the feature matrix X comprises:

[0167]

[0168] in represents the number of samples, m represents the number of features, and the elements in the matrix Representative The sample in The value of the feature, X m is the feature column of the mth feature in the feature matrix X, which contains the values ​​of all samples on the mth feature.

[0169] As the input layer of the neural network, the feature matrix unifies different types of features (continuous and discrete) into numerical form, which meets the input requirements of the neural network. Each sample is represented as a structured feature vector for easy batch processing.

[0170] The process of constructing the seabuckthorn full component extraction prediction model includes: model training, model verification and model testing;

[0171] The method for constructing the seabuckthorn full component extraction prediction model comprises:

[0172] A sample data set is obtained, wherein the sample data set includes a historical feature matrix, a historical extracted product type, and a historical extraction rate. The sample data set is divided into a sample training set, a sample verification set, and a sample test set. A neural network is constructed, and the historical feature matrix in the sample training set is used as input data of the neural network. The historical extracted product type and the historical extraction rate in the sample training set are used as output data of the neural network. The neural network is trained to obtain an initial neural network for predicting the extracted product type and the extraction rate. The sample verification set is used to adjust the parameters of the initial neural network, and the output of the neural network satisfies a preset error value as a sea buckthorn full ingredient extraction prediction model. Finally, the sea buckthorn full ingredient extraction prediction model is tested using the sample test set.

[0173] The model training process of the seabuckthorn full component extraction prediction model requires setting a loss function, and the method for obtaining the loss function of the seabuckthorn full component extraction prediction model includes:

[0174] For extracting product type prediction, this embodiment uses the cross entropy loss function;

[0175] For extraction rate prediction, this embodiment uses the mean square error loss function;

[0176] The weighted sum of the cross entropy loss function and the mean square error loss function was used as the loss function of the seabuckthorn full component extraction prediction model.

[0177] The weighted sum of the cross entropy loss function and the mean square error loss function can simultaneously minimize the prediction errors of the extracted product type and extraction rate.

[0178] The prediction model for seabuckthorn full component extraction adopts a multi-layer perceptron structure, such as Figure 2 As shown, it includes input layer, hidden layer and output layer;

[0179] The activation function in the hidden layer adopts one of the nonlinear activation functions Sigmoid, Tanh, and ReLU, preferably, the ReLU activation function is adopted;

[0180] The activation function introduces nonlinearity into the neural network and enhances the network's expressive power. Compared with other activation functions, the ReLU activation function has the advantages of fast convergence speed and unsaturated gradient.

[0181] The output layer includes output layer 1 and output layer 2;

[0182] The output layer 1 is used for prediction of extraction rate, which is a regression task. The output layer 1 includes 1 neuron, and the activation function is preferably an identity function;

[0183] The output layer 2 is used to predict and extract product types. The number of neurons in the output layer 2 is equal to the number of categories, and the activation function is preferably Softmax.

[0184] In this step, the multi-layer perceptron structure adds a hidden layer and introduces a nonlinear activation function, which makes the network have stronger fitting and expression capabilities; the gradient stability of the ReLU activation function is very good, which avoids the gradient vanishing problem and helps the network converge to the optimal solution quickly; the classification output layer (extracted product type output layer) and the regression output layer (extraction rate output layer) are designed in parallel, so that the model can simultaneously predict the extracted product type and extraction rate.

[0185] S60: Dynamically optimize the seabuckthorn full ingredient extraction process based on predicted extraction product types and extraction rates.

[0186] Step S60 in the embodiment of the present application also includes:

[0187] S61: Compare the predicted extracted product type and extraction rate with the target extracted product type and extraction rate, calculate the deviation, and obtain a product type prediction deviation vector and an extraction rate prediction deviation;

[0188]

[0189]

[0190] in, Extract the encoding vector of product type for the target, is the target extraction rate, is the product type prediction deviation vector, is the extraction rate prediction deviation, Encode the vector for the predicted extracted product type, is the predicted extraction rate; this formula quantifies the gap between the predicted result and the target value, providing a basis for subsequent process parameter regulation.

[0191] S62: Calculate the control amount of the process parameter based on the product type prediction deviation vector and the extraction rate prediction deviation;

[0192]

[0193]

[0194] in, For the current moment, is the control quantity vector of the process parameters at the current moment, is the comprehensive deviation at the current moment, For a moment in the past The combined deviation of is the weight matrix of product type prediction bias, Forecast deviation balance factor for product type, is the extraction rate prediction deviation balance factor, is the proportionality coefficient, is the integration coefficient, is the differential coefficient, is the prediction deviation vector of the product type at the current moment, is the extraction rate prediction deviation at the current moment; combining the type prediction deviation and the extraction rate prediction deviation, the control amount of the process parameters is calculated, where the comprehensive deviation is the weighted sum of the two deviations.

[0195] S63: performing decoupling calculation on the control amount of the process parameters to obtain the process parameter adjustment amount, applying the calculated process parameter adjustment amount to the actual production process, and dynamically optimizing the extraction process;

[0196] Considering the coupling relationship between different process parameters, the parameter coupling matrix R is introduced to decouple the control amount of the process parameters. The calculation formula is:

[0197]

[0198] in is the process parameter adjustment amount after decoupling at the current moment t;

[0199] Apply the calculated process parameter adjustments to the actual production process to dynamically optimize the extraction process:

[0200]

[0201] in, is the process parameter vector at the current moment, is the process parameter vector at the next moment.

[0202] The seabuckthorn full-ingredient extraction prediction model predicts the type of extraction products and extraction rate that can be obtained under the current process conditions. The process parameters can be intelligently adjusted and optimized accordingly, avoiding a large amount of manual experience and blind trial and error, reducing the time and experimental costs of process development; the process parameters are dynamically adjusted according to the real-time prediction results, so that the extraction process can be automated and intelligent, improving production efficiency and quality stability.

[0203] Example 2

[0204] This embodiment provides an intelligent optimization system for extracting all seabuckthorn ingredients based on the first embodiment. Figure 3 As shown, including:

[0205] Process feature set acquisition module: used to comprehensively collect various key features that affect seabuckthorn extraction results, including extraction methods, extraction conditions, pretreatment processes, seabuckthorn raw materials, seabuckthorn varieties and origins, etc., to lay a data foundation for subsequent analysis and avoid missing important information;

[0206] New feature construction module: used to obtain new features in the process parameters of seabuckthorn full-ingredient extraction, including pressure-temperature ratio, solvent polarity index and volume-surface area ratio, enrich the original feature space, provide more dimensional information for the model, and help improve prediction accuracy and generalization ability;

[0207] Data preprocessing module: used to process outliers, standardize and encode the feature set and new features of the seabuckthorn full component extraction process to improve data quality;

[0208] Seabuckthorn ingredient extraction prediction module: used to construct a feature matrix, and predict the extraction product type and extraction rate based on the feature matrix and the seabuckthorn full ingredient extraction prediction model;

[0209] Process parameter dynamic optimization module: It is used to dynamically optimize the seabuckthorn full-ingredient extraction process according to the predicted seabuckthorn extraction product type and extraction rate, so that the extraction process can be automated and intelligent, and the production efficiency and quality stability can be improved.

[0210] Example 3

[0211] This embodiment discloses an electronic device, including a memory, a central processing unit, and a computer program stored in the memory and executable on the central processing unit. When the central processing unit executes the computer program, the intelligent optimization method for extracting all seabuckthorn components provided above is implemented.

[0212] Since the electronic device introduced in this embodiment is an electronic device used to implement a method for intelligent optimization of seabuckthorn full-ingredient extraction process in the embodiment of this application, based on the method for intelligent optimization of seabuckthorn full-ingredient extraction process introduced in the embodiment of this application, a person skilled in the art can understand the specific implementation of the electronic device of this embodiment and its various variations, so how the electronic device implements the method in the embodiment of this application is not described in detail here. As long as a person skilled in the art implements the electronic device used in the method for intelligent optimization of seabuckthorn full-ingredient extraction process in the embodiment of this application, it belongs to the scope of protection of this application.

[0213] Example 4

[0214] This embodiment discloses a computer-readable storage medium, on which a computer program is stored. When the computer program is executed, the above-mentioned intelligent optimization method for extracting all the components of seabuckthorn is implemented.

[0215] The above formulas are all dimensionless and numerical calculations. The formula is a formula for the most recent real situation obtained by collecting a large amount of data and performing software simulation. The preset parameters, weights and thresholds in the formula are set by technicians in this field according to actual conditions.

[0216] The above embodiments may be implemented in whole or in part by software, hardware, firmware or any other combination thereof. When implemented by software, the above embodiments may be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer programs are loaded or executed on a computer, the process or function described in the embodiment of the present invention is generated in whole or in part. The computer may be a general-purpose computer, a special-purpose computer, a computer network or other programmable device. The computer instructions may be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another computer-readable storage medium. For example, the computer instructions may be transmitted from one website, computer, server or data center to another website, computer, server or data center via a wired network or a wireless network. The computer-readable storage medium may be any available medium that can be accessed by a computer or a data storage device such as a server or data center that includes one or more available media sets. The available medium may be a magnetic medium (e.g., a floppy disk, a hard disk, a tape), an optical medium (e.g., a DVD) or a semiconductor medium. The semiconductor medium may be a solid-state hard disk.

[0217] Those skilled in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed in the present invention can be implemented in electronic hardware or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of the present invention.

[0218] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the systems and units described above can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0219] In the several embodiments provided by the present invention, it should be understood that the disclosed systems and methods can be implemented in other ways. For example, the system embodiments described above are only schematic, for example, the division of the units is only one, and there may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be an indirect coupling or communication connection through some interfaces, devices or units, which can be electrical, mechanical or other forms.

[0220] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0221] In addition, each functional unit in each embodiment of the present invention may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit.

[0222] The above description is only a specific implementation mode of the present invention, but the protection scope of the present invention is not limited thereto. Any technician familiar with the technical field can easily think of changes or substitutions within the technical scope disclosed by the present invention, which should be covered by the protection scope of the present invention.

[0223] Finally: The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. An intelligent optimization method for extracting the whole components of seabuckthorn, characterized in that: include: S10: Obtaining a feature set of the seabuckthorn full component extraction process, wherein the feature set includes continuous features and discrete features; S20: constructing new features of the seabuckthorn full component extraction process, the new features including pressure-temperature ratio, solvent polarity index and volume-surface area ratio; S30: preprocessing the feature set and new features of the seabuckthorn full component extraction process; S40: coding of sea buckthorn varieties, origins, similarities between varieties, and associations between varieties and origins; S50: construct a feature matrix, and predict the extraction product type and extraction rate based on the feature matrix and the seabuckthorn full component extraction prediction model; S60: Dynamic optimization of seabuckthorn full ingredient extraction process based on predicted extraction product type and extraction rate; The method for encoding seabuckthorn varieties, origins, similarities between varieties, and associations between varieties and origins includes: constructing a seabuckthorn variety-origin heterogeneous graph, using a graph neural network algorithm to learn embedded representations of variety and origin nodes, and obtaining a final variety-origin representation; The method for constructing a seabuckthorn variety-origin heterogeneous graph comprises: obtaining variety-variety similarity edge information and obtaining variety-origin association edge information; The method for obtaining variety-variety similarity edge information includes: obtaining genome information of different seabuckthorn varieties, calculating the genetic distance between different seabuckthorn varieties according to the genome information, and obtaining the similarity between the varieties based on the genetic distance; The method for obtaining the variety-origin association edge information includes: counting the planting area proportion of each seabuckthorn variety in different origins, and constructing a A matrix A, where is the number of varieties, q is the number of origins, Indicates The species in The proportion of planting area in each production area; The method for calculating the genetic distance between different seabuckthorn varieties comprises: Assume that There are 100 varieties of sea buckthorn, and the genome of each variety is represented by a binary vector of length s, denoted as ,in, ; Indicates the variety There is a variation at the rth gene locus, Indicates no variation, variety and varieties The genetic distance between Use the following formula to measure: ； in represents the number of intersections between two seabuckthorn varieties at the mutation sites, represents the number of unions of two seabuckthorn varieties at the variant sites; The method for obtaining the similarity between varieties based on genetic distance comprises: ； in, For varieties and varieties The similarity between is the bandwidth parameter of the Gaussian kernel, is an exponential function; The method for dynamically optimizing the seabuckthorn full-ingredient extraction process based on predicted extraction product types and extraction rates comprises: comparing the predicted extraction product types and extraction rates with the target extraction product types and extraction rates, calculating deviations, and obtaining product type prediction deviation vectors and extraction rate prediction deviations; calculating the control amount of process parameters based on the product type prediction deviation vectors and extraction rate prediction deviations; performing decoupling calculation on the control amount of process parameters to obtain process parameter adjustment amounts, applying the calculated process parameter adjustment amounts to the actual production process, and dynamically optimizing the extraction process; The method of comparing the predicted extracted product type and extraction rate with the target extracted product type and extraction rate, calculating the deviation, and obtaining the product type prediction deviation vector and the extraction rate prediction deviation comprises: ； ； in, Extract the encoding vector of product type for the target, is the target extraction rate, is the product type prediction deviation vector, is the extraction rate prediction deviation, Encode the vector for the predicted extracted product type, is the predicted extraction rate; The method for calculating the control amount of the process parameter based on the product type prediction deviation vector and the extraction rate prediction deviation includes: ； ； in, For the current moment, is the control quantity vector of the process parameters at the current moment, The current moment is the comprehensive deviation, For a moment in the past The combined deviation of is the weight matrix of product type prediction bias, Forecast deviation balance factor for product type, is the extraction rate prediction deviation balance factor, is the proportionality coefficient, is the integration coefficient, is the differential coefficient, is the prediction deviation vector of the product type at the current moment, The predicted deviation of the extraction rate at the current moment.

2. The intelligent optimization method for extracting the full components of seabuckthorn according to claim 1, characterized in that: The process of preprocessing the feature set and new features of the seabuckthorn full component extraction process includes: outlier processing, standardization processing and category coding; the outlier processing method includes a weighted outlier processing method; the standardization processing method includes an adaptive standardization processing method; The weighted outlier processing method includes: calculating the weighted Euclidean distance of samples in the feature vector, calculating the mean and standard deviation of the weighted Euclidean distance of all samples, and judging and processing outliers; The method for calculating the weighted Euclidean distance of samples in the feature vector includes: ； in, is the weighted Euclidean distance of the i-th sample, x is the feature vector of the continuous feature, n is the number of samples, p is the number of features, is the kth eigenvalue of the ith sample, is the kth eigenvalue of the jth sample, For sample The weight of .

3. The intelligent optimization method for extracting the full components of seabuckthorn according to claim 2, characterized in that: The method for judging and processing abnormal values ​​comprises: ； in, is the mean of the sample weighted Euclidean distance, is the standard deviation of the sample weighted Euclidean distance, is the weighted Euclidean distance of the ith sample, z is the z-score value; set is the abnormal threshold; like ,but For outliers, after detecting them, they are removed or replaced with the average value of other samples; The adaptive normalization processing method comprises: ； ； in, is the adaptive parameter, is the kth eigenvalue of the ith sample, is the standardized value of the kth feature of the i-th sample, is the mean of the kth feature, is the standard deviation of the kth feature.

4. An intelligent optimization system for seabuckthorn full ingredient extraction process, which is used to implement an intelligent optimization method for seabuckthorn full ingredient extraction process according to any one of claims 1 to 3, characterized in that: include: Process feature set acquisition module: used to comprehensively collect various key features that affect the results of seabuckthorn extraction, including extraction methods, extraction conditions, pretreatment processes, seabuckthorn raw materials, seabuckthorn varieties and origins; New feature construction module: used to obtain new features in the process parameters of seabuckthorn full-component extraction, including pressure-temperature ratio, solvent polarity index and volume-surface area ratio; Data preprocessing module: used to process outliers, standardize and encode the feature set and new features of the seabuckthorn full component extraction process; Seabuckthorn ingredient extraction prediction module: used to construct a feature matrix, and predict the extraction product type and extraction rate based on the feature matrix and the seabuckthorn full ingredient extraction prediction model; Process parameter dynamic optimization module: used to dynamically optimize the seabuckthorn full ingredient extraction process according to the predicted seabuckthorn extraction product type and extraction rate.

5. An electronic device comprising a memory, a central processing unit, and a computer program stored in the memory and executable on the central processing unit, characterized in that: When the central processing unit executes the computer program, an intelligent optimization method for extracting the full components of seabuckthorn according to any one of claims 1 to 3 is implemented.

6. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, and when the computer program is executed, the method for intelligent optimization of the seabuckthorn full-ingredient extraction process described in any one of claims 1 to 3 is implemented.

Images