An intelligent temperature control method for supercritical extraction of plant essential oils

Abstract

The present invention relates to the field of industrial monitoring and scheduling, and particularly to an intelligent temperature control method for supercritical extraction of plant essential oils. The method clusters the historical temperature monitoring time series data window in the extraction kettle during the extraction process of plant essential oils to obtain the hidden states corresponding to the historical temperature monitoring time series data; establishes a hidden Markov prediction model according to the hidden states of the historical temperature monitoring time series data of the extraction kettle, and predicts the real-time temperature monitoring time series data of the extraction kettle according to the hidden Markov prediction model to obtain the state transition probability matrix of the real-time temperature monitoring data; obtains the parameter optimization factor of the real-time temperature monitoring data according to the state transition probability matrix of the real-time temperature monitoring data; and obtains a PID control model for real-time temperature control of the extraction kettle according to the parameter optimization factor, which can obtain more accurate parameters of the temperature control model of the extraction kettle and improve the accuracy of the temperature control model during the extraction process of plant essential oils.

Description

Technical Field

[0001] The present invention relates to the technical field of industrial monitoring and scheduling, and particularly to an intelligent temperature control method for supercritical extraction of plant essential oils. Background Art

[0002] Carbon dioxide supercritical extraction technology is an efficient and environmentally friendly separation method, which is widely used in the field of extraction of plant essential oils. Its principle is based on the unique property that carbon dioxide has both the diffusivity of a gas and the solubility of a liquid in the supercritical state. By controlling the temperature and pressure, carbon dioxide in the supercritical state is brought into contact with plant materials, dissolving the essential oil components therein, and the essential oil is separated and collected in the separation kettle by reducing the pressure or lowering the temperature. This technology has significant advantages such as no solvent residue, high extraction efficiency, mild operation and strong selectivity, and can effectively extract heat-sensitive components and ensure product purity. However, in practical applications, challenges such as the control accuracy of temperature and pressure, equipment energy consumption, and optimization of process parameters for different raw materials need to be faced.

[0003] Regarding the temperature control of the extraction kettle for supercritical extraction of plant essential oils, in the prior art, after all historical temperature monitoring time series data of the plant essential oil extraction kettle are obtained, the parameters of the PID control model are fitted according to the output power time series data of the heating device during the plant essential oil extraction process, and the parameters of the PID control model are obtained through iterative optimization according to the real-time temperature monitoring time series data, so as to obtain the PID model for temperature control of the plant essential oil extraction kettle.

[0004] Among them, in the process of controlling the temperature of the plant essential oil extraction kettle through the PID model, due to the possible fluctuations in the temperature of the extraction kettle, when controlling the temperature solely through the PID control model, overshoot or oscillation may occur in the control of the temperature in the extraction kettle when the temperature of the extraction kettle fluctuates.

[0005] Therefore, how to ensure the accuracy of temperature control during the supercritical extraction of plant essential oils and thus improve the stability of plant essential oil extraction has become an urgent problem to be solved. Summary of the Invention

[0006] In view of this, an embodiment of the present invention provides an intelligent temperature control method for supercritical extraction of plant essential oils to solve the problem of how to ensure the accuracy of temperature control under temperature fluctuation conditions during the temperature control of the extraction kettle, and thus improve the temperature control accuracy of the extraction kettle.

[0007] An embodiment of the present invention provides an intelligent temperature control method for supercritical extraction of plant essential oils. The intelligent temperature control method for supercritical extraction of plant essential oils includes the following steps:

[0008] Collect the time-series data of temperature monitoring in the supercritical extraction kettle of plant essential oil, obtain the historical temperature monitoring time-series data of the extraction kettle, divide the window of all historical temperature monitoring time-series data, and obtain all temperature time-series windows;

[0009] For any extraction kettle, perform clustering analysis according to the optimized distance metric between the historical temperature monitoring time-series data windows to obtain the cluster division result; according to the cluster division result of the historical temperature monitoring time-series data windows, obtain the hidden state of each historical temperature monitoring time-series data window;

[0010] Establish a hidden Markov prediction model according to the historical temperature monitoring time-series data and the hidden state of the historical temperature monitoring time-series data windows; predict through the hidden Markov prediction model according to the real-time temperature monitoring time-series data of the extraction kettle to obtain the state transition probability matrix of the real-time temperature monitoring time-series data of the extraction kettle; according to the state transition probability matrix and the state fluctuation evaluation of the clusters in the cluster division result, obtain the parameter optimization factor of the real-time temperature monitoring time-series data of the extraction kettle;

[0011] Obtain a PID control model for the real-time temperature control of the extraction kettle according to the historical temperature monitoring time-series data of the extraction kettle, including the parameters after convergence of the PID control model; optimize the parameters of the PID control model for the real-time temperature monitoring data of the extraction kettle according to the parameter optimization factor of the real-time temperature monitoring time-series data of the extraction kettle, and perform temperature control for the supercritical extraction of plant essential oil according to the PID control model with optimized parameters.

[0012] Preferably, the performing clustering analysis according to the optimized distance metric between the historical temperature monitoring time-series data windows to obtain the cluster division result includes:

[0013] For any historical temperature monitoring time-series data window, obtain the timestamp distance between the data points in the historical temperature monitoring time-series data window and the window center point, take the reciprocal of the result of adding it to the constant 1 as the first influence parameter, and obtain the outlier detection of the data points in the historical temperature monitoring time-series data window to obtain the outlier factor of the data in the historical temperature monitoring time-series data window, and normalize the absolute value result of subtracting the outlier factor from the constant 1 as the second influence parameter. Multiply the first influence parameter by the result of the constant 1 and the second influence parameter to obtain the result after normalization as the contribution degree of fluctuation analysis corresponding to the data in the historical temperature time-series data window;

[0014] For any of the historical temperature monitoring time series data windows, based on the contribution degree of the data in the historical temperature monitoring time series data window to the fluctuation analysis, the historical temperature monitoring time series data window, and the temperature mean value of the historical temperature monitoring time series data window, the difference between the data in the historical temperature monitoring time series data window and the temperature mean value of the historical temperature monitoring time series data window is used as the first fluctuation measure. Based on the contribution degree of the data corresponding to the fluctuation analysis in the historical temperature time series data window, all the first fluctuation measures are weighted and summed and divided by the window length of the historical temperature time series data window to obtain the corresponding fluctuation calculation result;

[0015] Obtain the slope in the least squares fitting result of the historical temperature time series data window, and take the absolute value of the difference between the fluctuation calculation result of the historical temperature monitoring time series data window and the fluctuation calculation result of the historical temperature monitoring time series data window corresponding to the cluster center in the clustering process as the first measure of the difference optimization factor. Take the absolute value of the difference between the slope in the least squares fitting result of the historical temperature time series data window and the slope in the least squares fitting result of the historical temperature monitoring time series data window corresponding to the cluster center in the clustering process as the second measure of the difference optimization factor; Normalize the product of the first measure of the difference optimization factor and the second measure of the difference optimization factor to obtain the corresponding normalized value as the difference optimization factor between the historical temperature monitoring time series data window and the historical temperature monitoring time series data window corresponding to the cluster center in the clustering process;

[0016] For any of the historical temperature monitoring time series data windows in the clustering process, take the Euclidean distance between the historical temperature monitoring time series data window and the historical temperature monitoring time series data window corresponding to the cluster center in the clustering process as the first Euclidean distance; Take the product of the difference optimization factor between the historical temperature monitoring time series data window and the historical temperature monitoring time series data window corresponding to the cluster center in the clustering process and the first Euclidean distance as the optimized distance measure between the historical temperature monitoring time series data window and the historical temperature monitoring time series data window corresponding to the cluster center in the clustering process;

[0017] Based on the optimized distance measure between the historical temperature monitoring time series data window and the historical temperature monitoring time series data window corresponding to the cluster center in the clustering process through The clustering algorithm completes the clustering process, and obtains the cluster division result of the historical temperature monitoring time series data window.

[0018] Preferably, obtaining the hidden state of each historical temperature monitoring time series data window according to the cluster division result of the historical temperature monitoring time series data window includes:

[0019] Obtain the clustering result of the historical temperature monitoring time series data window, and use the cluster number of the historical temperature monitoring time series data window as the hidden state of each historical temperature monitoring time series data window.

[0020] Preferably, the method for obtaining the state transition probability matrix of the real-time temperature monitoring time series data of the extraction kettle according to the hidden Markov prediction model of the historical temperature monitoring time series data includes:

[0021] Obtain the hidden Markov model of the historical temperature monitoring time series data, and obtain the real-time temperature monitoring time series data of the extraction kettle. According to the hidden Markov model of the historical temperature monitoring time series data, obtain the state transition probability matrix corresponding to the real-time temperature monitoring time series data of the extraction kettle.

[0022] Preferably, the method for obtaining the parameter optimization factor of the real-time temperature monitoring time series data of the extraction kettle according to the state transition probability matrix and the state fluctuation evaluation of the clusters in the clustering result includes:

[0023] Obtain the state transition probability matrix of the real-time temperature monitoring time series data of the extraction kettle, and use the state transition probability of the target state in the state transition probability matrix of the real-time temperature monitoring time series data of the extraction kettle as the fluctuation weight; obtain the clustering result of the historical temperature monitoring time series data window in the clustering process, and according to the first fluctuation measure of the historical temperature monitoring time series data window in each cluster, obtain the state fluctuation evaluation of each cluster in the clustering result; perform weighted summation on the state fluctuation evaluations of all the clusters with the fluctuation weight of the real-time temperature monitoring time series data of the extraction kettle to obtain the corresponding weighted summation result; substitute the opposite number of the weighted summation result into the exponential function with the natural constant e as the base, and the corresponding first exponential function result is obtained. Use the subtraction result between the constant 1 and the first exponential function result as the parameter optimization factor of the real-time temperature monitoring time series data of the extraction kettle.

[0024] Preferably, the method for obtaining the state fluctuation evaluation of each cluster in the clustering result includes:

[0025] Obtain the clustering result of the historical temperature monitoring time series data window, and use the mean value of the fluctuation calculation of the historical temperature monitoring time series data window in each cluster in the clustering result of the historical temperature monitoring time series data window as the state fluctuation evaluation of the cluster.

[0026] Preferably, the method for optimizing the parameters of the PID control model of the real-time temperature monitoring data of the extraction kettle according to the parameter optimization factor of the real-time temperature monitoring time series data of the extraction kettle to obtain the PID control model of the real-time temperature monitoring time series data of the extraction kettle includes:

[0027] Obtain the integral gain parameter of the real-time temperature PID control model of the extraction kettle and the parameter optimization factor of the real-time temperature monitoring time series data of the extraction kettle respectively; add the constant 1 to the parameter optimization factor of the real-time temperature monitoring time series data of the extraction kettle as the first parameter weight, and weight the integral gain parameter of the real-time temperature PID control model of the extraction kettle according to the first parameter weight, and take the result as the optimized integral gain parameter.

[0028] Take the optimized integral gain parameter as the real-time integral gain parameter in the real-time temperature PID control model of the extraction kettle, and obtain the PID control model of the real-time temperature control time series data of the extraction kettle.

[0029] Preferably, window partitioning is performed on all historical temperature monitoring time series data to obtain all temperature time series windows, including:

[0030] Obtain the historical temperature monitoring time series data of the extraction kettle and the preset window length respectively; perform a traversal with a continuous step size of 1 on the historical temperature monitoring time series data of the extraction kettle according to the preset window length, and use the subsequence included in each movement of the window with the predicted window length as the historical temperature monitoring time series data window corresponding to the center point of the window.

[0031] The beneficial effects of the embodiments of the present invention compared with the prior art are:

[0032] Collect the time-series data of temperature monitoring in the supercritical extraction kettle of plant essential oil, obtain the historical time-series data of temperature monitoring of the extraction kettle, divide the window of all historical time-series data of temperature monitoring, and obtain all temperature time-series windows; for any extraction kettle, perform clustering analysis according to the optimized distance metric between the historical time-series data windows of temperature monitoring, and obtain the clustering result of cluster classes. According to the clustering result of the historical time-series data windows of temperature monitoring, obtain the hidden state of each historical time-series data window of temperature monitoring; according to the historical time-series data of temperature monitoring and the hidden state of the historical time-series data windows of temperature monitoring, establish a hidden Markov prediction model; predict through the hidden Markov prediction model according to the real-time temperature monitoring time-series data of the extraction kettle, and obtain the state transition probability matrix of the real-time temperature monitoring time-series data of the extraction kettle; according to the state transition probability matrix and the state fluctuation evaluation of the clusters in the clustering result, obtain the parameter optimization factor of the real-time temperature monitoring time-series data of the extraction kettle; according to the historical time-series data of temperature monitoring of the extraction kettle, obtain the PID control model for the real-time temperature control of the extraction kettle, including the parameters after convergence of the PID control model; according to the parameter optimization factor of the real-time temperature monitoring time-series data of the extraction kettle, optimize the parameters of the PID control model for the real-time temperature monitoring data of the extraction kettle, and perform temperature control for the supercritical extraction of plant essential oil according to the PID control model with optimized parameters. Among them, by performing fluctuation analysis on all historical time-series data windows of temperature monitoring of the extraction kettle, the optimized distance metric between the historical time-series data windows of temperature monitoring is obtained, clustering is performed through this optimized distance metric, the clustering result is obtained, each cluster class is used as a temperature fluctuation state, used to determine the hidden state of each temperature time-series data window, a hidden Markov prediction model is established through the temperature monitoring time-series data and the hidden state of the temperature time-series data window, the real-time temperature monitoring time-series data of the extraction kettle is used to obtain the state transition probability matrix through the hidden Markov prediction model, and the parameter optimization factor is obtained through the state transition probability matrix and the state fluctuation evaluation of each state, used to optimize the integral gain parameter of the PID model when regulating the real-time temperature of the extraction kettle through the PID model, so as to obtain more rigorous PID model parameters, thus avoiding the situation of temperature overshoot or oscillation during the temperature regulation process and improving the control accuracy of the temperature control model. Description of the Drawings

[0033] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following will briefly introduce the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings in the following description are only some embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0034] Figure 1 It is a flowchart of a method for intelligent temperature control in the supercritical extraction of plant essential oils provided by Embodiment 1 of the present invention. Specific embodiments

[0035] The embodiments of the present disclosure will be described in detail below. The examples of the embodiments are shown in the accompanying drawings, where the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and are intended to explain the present disclosure, and should not be construed as limiting the present disclosure.

[0036] It should be noted that the terms "first", "second", etc. in the description of the present disclosure and the above accompanying drawings are used to distinguish similar objects and do not necessarily have to be used to describe a specific order or sequence. It should be understood that such data can be interchanged under appropriate circumstances so that the embodiments of the present disclosure described herein can be implemented in an order other than those illustrated or described herein. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with the present disclosure. On the contrary, they are merely examples of devices and methods consistent with some aspects of the present disclosure.

[0037] To illustrate the technical solution of the present invention, it will be described below through specific embodiments.

[0038] The specific scenario targeted by the present invention is: intelligent temperature control of the extraction kettle during the supercritical extraction of plant essential oils by carbon dioxide.

[0039] See Figure 1 , which is a flowchart of a method for intelligent temperature control in the supercritical extraction of plant essential oils provided by Embodiment 1 of the present invention. As Figure 1 shown, a method for intelligent temperature control in the supercritical extraction of plants may include:

[0040] Step S101, collect the temperature monitoring time series data in the supercritical extraction kettle of plant essential oils, obtain the historical temperature monitoring time series data of the extraction kettle, and divide all the historical temperature monitoring time series data into windows to obtain all temperature time series windows.

[0041] For the supercritical extraction process of plant essential oils, all historical temperature monitoring time series data in the extraction kettle can be collected through the temperature sensor in the extraction kettle. Preferably, in the embodiment of the present invention, the sampling frequency of the temperature sensor in the extraction kettle is preset to seconds for one collection, that is, the temperature of the extraction kettle is recorded every seconds and the temperature time series data is formed as the corresponding temperature monitoring time series data of the extraction kettle.

[0042] Based on the above data acquisition method, historical temperature monitoring time series data of the extraction kettle is obtained. After obtaining the historical temperature monitoring time series data of the extraction kettle, it is necessary to perform window division on the historical temperature monitoring time series data according to the preset time series window length. Preferably, in the embodiment of the present invention, the window length is set , starting from the th data point in the historical temperature time series data of the extraction kettle, and traversing all the historical temperature monitoring time series data of the extraction kettle with a window of length with a step size of . Each time the window moves, the data within the window is used as the temperature time series data window corresponding to the center point of the current window.

[0043] Step S102, for any extraction kettle, perform clustering analysis according to the optimized distance metric between the historical temperature monitoring time series data windows to obtain a cluster class division result. According to the cluster class division result of the historical temperature monitoring time series data windows, obtain the hidden state of each historical temperature monitoring time series data window.

[0044] After obtaining the historical temperature monitoring time series data window of the extraction kettle, the cluster division of all historical temperature monitoring time series data windows of the extraction kettle can be carried out through the clustering process, so as to obtain the hidden state of the historical temperature monitoring time series data window. Considering that in the clustering process of the historical temperature monitoring time series data window of the extraction kettle, the Euclidean distance is used to measure the distance between temperature monitoring time series data windows. However, in the historical temperature monitoring time series data of the extraction kettle, there are cases of different temperature change states. The different temperature change states refer to the cases of temperature rising, falling and continuous fluctuation in the historical temperature monitoring time series data of the extraction kettle. When measuring the distance between temperature time series data windows in the clustering process, there will be temperature time series data windows with different change patterns but similar distances. For example, in the first temperature time series data window, the temperature change shows a slow upward trend, in the second temperature time series data window, the temperature change shows a slow downward trend, and in the third window, the temperature shows a state of frequent jitter. The result of measuring the distance between the first temperature time series data window and the second temperature time series data window by the Euclidean distance may be similar to the result of measuring the distance between the first temperature time series data window and the third temperature time series data window by the Euclidean distance. Therefore, for the distance measurement between temperature time series data windows, it is necessary to evaluate through the characteristics presented by the different fluctuation states between two temperature time series data windows, so as to optimize the distance measurement between temperature time series data windows through the difference in the fluctuation characteristics presented between windows. At the same time, considering the possibility of abnormal temperature monitoring data in the temperature time series data window, and in the distance measurement between the historical temperature monitoring time series data windows of the extraction kettle, because there is a transition stage between different fluctuation modes in the temperature change of the extraction kettle, so in the process of measuring the distance between the historical temperature monitoring time series data windows of the extraction kettle, if the same weight is used for each data point in the time series data window to measure the distance between windows, it will lead to inaccurate cluster division of the data windows in the transition stage. Therefore, for the distance measurement between historical temperature monitoring time series data windows, it is necessary to optimize the evaluation of the difference in fluctuation characteristics through the evaluation of the contribution degree of the fluctuation analysis of the temperature data in the historical temperature time series data window. Therefore, in the embodiment of the present invention, first, according to the timestamp difference between the temperature data in any historical temperature monitoring time series data window and the outlier degree of the data point itself, the contribution degree of the fluctuation analysis corresponding to the data in the historical temperature time series data window is obtained, including:

[0045] For any historical temperature monitoring time series data window, obtain the timestamp distance between the data point in the historical temperature monitoring time series data window and the window center point, and take the reciprocal of the sum of the result and the constant 1 as the first influence parameter, and obtain the Outlier detection is performed to obtain the outlier factor of the data within the historical temperature monitoring time series data window. The absolute value result of subtracting the outlier factor from the constant 1 is normalized as the second influence parameter. Multiply the first influence parameter by the result of the constant 1 minus the second influence parameter for The normalized result is used as the contribution degree of the data corresponding to the historical temperature time series data window in the fluctuation analysis.

[0046] In one embodiment, for the historical temperature monitoring time series data window of the th extraction kettle, the calculation expression for the contribution degree of the

[0047]

[0048] th temperature data is as follows: where represents the contribution degree of the th temperature data in the th historical temperature monitoring time series data window in the fluctuation analysis, represents the normalization function, represents the absolute value symbol, represents a constant, represents the th historical temperature monitoring time series data window, represents the number of timestamp intervals between the th temperature data and the center point of the window, represents the th temperature data in the distance neighborhood of for the outlier factor, The size of the distance neighborhood can be adjusted according to the real-time scenario and is not required.

[0049] It should be noted that during the distance measurement process between temperature time series data windows, the larger the timestamp interval between the th temperature data in the data window and the center point of the window, the farther the position of the temperature data in the window is from the center point of the window, and the smaller its contribution degree to the center point of the window in the fluctuation analysis, and the corresponding value is smaller; at the same time, during the distance measurement process between temperature time series data windows, the th temperature data in the data window the absolute value of the difference between the outlier factor and the constant 1 The larger it is, the higher the degree of abnormality of the temperature data, and the smaller the contribution degree of the temperature data in the process of fluctuation analysis, so as to avoid the influence of abnormal data on the temperature time series data window in the process of fluctuation analysis.

[0050] After determining the contribution degree of each temperature data in the historical temperature time series data window to the fluctuation analysis, according to the contribution degree of the data in the historical temperature monitoring time series data window to the fluctuation analysis, the historical temperature monitoring time series data window and the temperature average value of the historical temperature monitoring time series data window, obtain the difference optimization factor between the historical temperature monitoring time series data window and the historical temperature monitoring time series data window corresponding to the cluster center in the clustering process, including:

[0051] Take the difference between the data in the historical temperature monitoring time series data window and the temperature average value of the historical temperature monitoring time series data window as the first fluctuation measure. According to the contribution degree of the data corresponding to the historical temperature time series data window to the fluctuation analysis, perform a weighted sum of all the first fluctuation measures and divide by the window length of the historical temperature time series data window to obtain the corresponding fluctuation calculation result;

[0052] Obtain the slope in the least squares fitting result of the historical temperature time series data window, and take the absolute value of the difference between the fluctuation calculation result of the historical temperature monitoring time series data window and the fluctuation calculation result of the historical temperature monitoring time series data window corresponding to the cluster center in the clustering process as the first measure of the difference optimization factor. Take the absolute value of the difference between the slope in the least squares fitting result of the historical temperature time series data window and the slope in the least squares fitting result of the historical temperature monitoring time series data window corresponding to the cluster center in the clustering process as the second measure of the difference optimization factor. Normalize the product of the first measure of the difference optimization factor and the second measure of the difference optimization factor to obtain the corresponding normalized value as the difference optimization factor between the historical temperature monitoring time series data window and the historical temperature monitoring time series data window corresponding to the cluster center in the clustering process.

[0053] In an embodiment, assume that in the clustering process, the th temperature data in the th historical temperature monitoring time series data window is , and the th temperature data in the historical temperature monitoring time series data window corresponding to the cluster center point of the th cluster is , then the calculation expression corresponding to the difference optimization factor between the th historical temperature monitoring time series data window and the historical temperature monitoring time series data window corresponding to the cluster center point of the th cluster is:

[0054]

[0055] Among them, represents the difference optimization factor between the th historical temperature monitoring time series data window and the historical temperature monitoring time series data window corresponding to the cluster center point of the th cluster class. , respectively represent the contribution degree of the fluctuation analysis of the th temperature data in the th historical temperature monitoring time series data window and the contribution degree of the fluctuation analysis of the th temperature data in the historical temperature monitoring time series data window corresponding to the cluster center point of the th cluster class. respectively represent the temperature mean value in the th historical temperature monitoring time series data window and the temperature mean value in the historical temperature monitoring time series data window corresponding to the cluster center point of the th cluster class. represents the window length. respectively represent the fitting slopes obtained by least squares fitting of the time series data in the historical temperature monitoring time series data window and the historical temperature monitoring time series data window corresponding to the cluster center point of the th cluster class. represents the linear normalization function. represents the absolute value symbol. represents the difference in the weighted variance of the contribution degree of the fluctuation analysis between two windows, that is, the first measure of the difference optimization factor. represents the difference in the fitting slopes between two windows, that is, the second measure of the difference optimization factor.

[0056] It should be noted that in the distance measurement between historical temperature monitoring time series data windows during the clustering process, the fluctuation difference between windows needs to be measured by the difference between the fluctuation calculation results between windows, so as to distinguish different fluctuation amplitudes between windows. Therefore, the difference in the fluctuation degree is analyzed through the first measure of the difference optimization factor. The greater the difference in the fluctuation calculation results between two windows, the greater the difference optimization factor of the corresponding two windows; at the same time, because there are also differences in the fluctuation trends between windows, the difference in the fluctuation trends is analyzed through the second measure of the difference optimization factor. The greater the difference in the fluctuation trends between two windows, the greater the difference optimization factor of the corresponding two windows.

[0057] After obtaining the difference optimization factor between the historical temperature monitoring time series data window and the historical temperature monitoring time series data window corresponding to the cluster center in the clustering process, according to the difference optimization factor between the historical temperature monitoring time series data window and the historical temperature monitoring time series data window corresponding to the cluster center in the clustering process, obtain the optimized distance metric between the historical temperature monitoring time series data window and the historical temperature monitoring time series data window corresponding to the cluster center in the clustering process, including:

[0058] For any historical temperature monitoring time series data window in the clustering process, the Euclidean distance between the historical temperature monitoring time series data window and the historical temperature monitoring time series data window corresponding to the cluster center in the clustering process is used as the first Euclidean distance. The product of the difference optimization factor between the historical temperature monitoring time series data window and the historical temperature monitoring time series data window corresponding to the cluster center in the clustering process and the first Euclidean distance is used as the optimized distance metric between the historical temperature monitoring time series data window and the historical temperature monitoring time series data window corresponding to the cluster center in the clustering process.

[0059] In one embodiment, taking the historical temperature time series data window corresponding to the cluster center of the th cluster as an example, the calculation expression of the distance metric between the th historical temperature monitoring time series data window and the historical temperature time series data window corresponding to the cluster center of the th cluster is:

[0060]

[0061] Wherein, represents the distance metric between the th historical temperature monitoring time series data window and the historical temperature time series data window corresponding to the cluster center of the th cluster, represents the first Euclidean distance between the th historical temperature monitoring time series data window and the historical temperature time series data window corresponding to the cluster center of the th cluster, represents the difference optimization factor between the th historical temperature monitoring time series data window and the historical temperature monitoring time series data window corresponding to the cluster center point of the th cluster.

[0062] After obtaining the optimized distance metric between the historical temperature monitoring time series data window and the historical temperature monitoring time series data window corresponding to the cluster center in the clustering process, according to the optimized distance metric between the historical temperature monitoring time series data window and the historical temperature monitoring time series data window corresponding to the cluster center in the clustering process, through The clustering algorithm completes the clustering process, and obtains the cluster division result of the historical temperature monitoring time series data window.

[0063] It should be noted that in one embodiment, after obtaining the optimized distance metric between the historical temperature monitoring time series data window and the historical temperature monitoring time series data window corresponding to the cluster center in the clustering process, through of Clustering completes the clustering process. The number of clusters in the clustering can be adjusted according to the real-time scenario and there is no requirement.

[0064] After obtaining the cluster division result of the historical temperature monitoring time series data window, according to the cluster division result of the historical temperature monitoring time series data window, obtain the hidden state of each historical temperature monitoring time series data window, including:

[0065] Obtain the cluster division result of the historical temperature monitoring time series data window, and use the number of the cluster of the historical temperature monitoring time series data window as the hidden state of each historical temperature monitoring time series data window.

[0066] Step S103, according to the historical temperature monitoring time series data and the hidden state of the historical temperature monitoring time series data window, establish a hidden Markov prediction model, and perform prediction on the real-time temperature monitoring time series data of the extraction kettle through the hidden Markov prediction model to obtain the state transition probability matrix of the real-time temperature monitoring time series data of the extraction kettle. According to the state transition probability matrix and the state fluctuation evaluation of the clusters in the cluster division result, obtain the parameter optimization factor of the real-time temperature monitoring time series data of the extraction kettle.

[0067] After obtaining the hidden state of each historical temperature monitoring time series data window, according to the historical temperature monitoring time series data and the hidden state of the historical temperature monitoring time series data window, establish a hidden Markov prediction model, and perform prediction on the real-time temperature monitoring time series data of the extraction kettle through the hidden Markov model to obtain the state transition probability matrix of the real-time temperature monitoring time series data of the extraction kettle. The establishment of the hidden Markov prediction model and the acquisition of the state transition probability matrix through the Viterbi algorithm are both well-known technologies and will not be elaborated in this embodiment.

[0068] After obtaining the state transition probability matrix of the real-time temperature monitoring time series data of the extraction kettle, according to the clustering result, obtain the state fluctuation evaluation of each cluster in the clustering result, including:

[0069] Obtain the clustering result of the historical temperature monitoring time series data window, and use the mean value of the fluctuation calculation of the historical temperature monitoring time series data window in each cluster in the clustering result of the historical temperature monitoring time series data window as the state fluctuation evaluation of the cluster.

[0070] In one embodiment, taking the th cluster as an example, the calculation expression for obtaining the state fluctuation evaluation of the th cluster is:

[0071]

[0072] Where represents the state fluctuation evaluation of the th cluster, represents the contribution degree of the fluctuation analysis of the th temperature data in the th historical temperature monitoring time series data window, represents that the th temperature data in the th historical temperature monitoring time series data window is , represents the temperature mean value in the th historical temperature monitoring time series data window, represents the length of the time series data window, represents the th number of historical temperature time series monitoring data windows in the cluster, and 1 represents the constant 1.

[0073] It should be noted that in the process of evaluating the state fluctuation of the cluster, using the mean value of the fluctuation calculation results of all historical temperature monitoring time series data windows in the cluster for the state fluctuation evaluation of the cluster can characterize the corresponding fluctuation mode through the overall state of the historical temperature monitoring time series data in the cluster. The higher the average fluctuation of the historical temperature monitoring time series data window in the cluster, the higher the state fluctuation evaluation of the cluster, and the larger the corresponding value, the higher the state fluctuation evaluation of the cluster.

[0074] After obtaining the state fluctuation evaluation of each cluster, according to the state transition probability matrix and the state fluctuation evaluation of the cluster in the clustering result, obtain the parameter optimization factor of the real-time temperature monitoring time series data of the extraction kettle, including:

[0075] Obtain the state transition probability matrix of the real-time temperature monitoring time series data of the extraction kettle, and use the state transition probability of the target state in the state transition probability matrix of the real-time temperature monitoring time series data of the extraction kettle as the fluctuation weight; obtain the cluster division result of the historical temperature monitoring time series data window of the clustering process, and according to the first fluctuation measure of the historical temperature monitoring time series data window in each cluster, obtain the state fluctuation evaluation of each cluster in the cluster division result; perform a weighted sum of the state fluctuation evaluations of all the clusters with the fluctuation weight of the real-time temperature monitoring time series data of the extraction kettle to obtain the corresponding weighted sum result; substitute the negative value of the weighted sum result into the exponential function with the natural constant e as the base, and the corresponding first exponential function result, and use the subtraction result between the constant 1 and the first exponential function result as the parameter optimization factor of the real-time temperature monitoring time series data of the extraction kettle.

[0076] In one embodiment, obtain the real-time temperature monitoring time series data window of the extraction kettle, denoted as , then the calculation expression of the parameter optimization factor of the real-time temperature monitoring time series data of the extraction kettle is:

[0077]

[0078] Where, represents the parameter optimization factor of the real-time temperature monitoring time series data of the extraction kettle, represents the probability that the real-time temperature monitoring time series data of the extraction kettle transfers to the hidden state corresponding to the th cluster in the state transition probability matrix, represents the th cluster's state fluctuation evaluation, represents the exponential function with the natural constant e as the base, 1 represents the constant 1, represents the number of clusters in the clustering process.

[0079] It should be noted that the parameter optimization factor is measured based on the state transition probability matrix corresponding to the real-time temperature monitoring data of the extraction kettle and the state fluctuation evaluations of all clusters. By performing a weighted sum of the state fluctuation evaluations corresponding to the states with the transition probability of the target state, when there is a high probability of transferring to a high-fluctuation state, the corresponding value is larger, and the parameter optimization factor of the real-time temperature monitoring time series data of the extraction kettle is larger.

[0080] Step S104: Obtain a PID control model for real-time temperature control of the extraction kettle according to the historical temperature monitoring time-series data of the extraction kettle, including the parameters after convergence of the PID control model; optimize the parameters of the PID control model for the real-time temperature monitoring data of the extraction kettle according to the parameter optimization factor of the real-time temperature monitoring time-series data of the extraction kettle, and perform temperature control for the supercritical extraction of plant essential oils according to the PID control model with optimized parameters.

[0081] After obtaining the parameter optimization factor of the real-time temperature monitoring time-series data of the extraction kettle, use the parameter optimization factor to optimize the integral gain parameter in the PID control model for real-time temperature control of the extraction kettle temperature, so as to obtain an accurate extraction kettle temperature control model. Therefore, in the embodiment of the present invention, according to the historical temperature monitoring time-series data of the extraction kettle, obtain a PID control model for real-time temperature control of the extraction kettle temperature, including the parameters after convergence of the PID control model; optimize the parameters of the PID control model for the real-time temperature monitoring data of the extraction kettle according to the parameter optimization factor of the real-time temperature monitoring time-series data of the extraction kettle, and perform temperature control for the supercritical extraction of plant essential oils according to the PID control model with optimized parameters, including:

[0082] Respectively obtain the integral gain parameter of the real-time temperature PID control model of the extraction kettle and the parameter optimization factor of the real-time temperature monitoring time-series data of the extraction kettle; add the constant 1 to the parameter optimization factor of the real-time temperature monitoring time-series data of the extraction kettle as the first parameter weight, and weight the integral gain parameter of the real-time temperature PID control model of the extraction kettle according to the first parameter weight, and use the result as the optimized integral gain parameter; use the optimized integral gain parameter as the real-time integral gain parameter in the real-time temperature PID control model of the extraction kettle to obtain the PID control model with optimized parameters.

[0083] In an embodiment, assume that the initial integral gain parameter in the real-time temperature PID control model of the extraction kettle is , then the calculation expression for real-time optimization of the integral gain parameter by the parameter optimization factor of the real-time temperature monitoring time-series data window of the extraction kettle is:

[0084]

[0085] Among them, represents the real-time integral gain parameter of the real-time temperature PID control model of the extraction kettle, represents the parameter optimization factor of the real-time temperature monitoring time-series data of the extraction kettle, 1 represents the constant 1, represents the initial integral gain parameter in the real-time temperature PID control model of the extraction kettle.

[0086] It should be noted that the real-time temperature monitoring time series data window in the extraction kettle in the embodiment of the present invention uses the window of the temperature data before the latest temperature monitoring time series data collected by the temperature sensor in the extraction kettle as the real-time data window, so as to ensure the integrity of the temperature data in the temperature time series data window. In the PID model used for the temperature control of the extraction kettle, its initial parameters are iterated through the regulation process of the historical temperature monitoring time series data until the PID model parameters converge, and the converged PID control model parameters are used as the initial parameters. The establishment and parameter acquisition of the PID model belong to the prior art and will not be elaborated here. In the PID control model, the higher the integral gain parameter, the lower the sensitivity of the control system. Therefore, in the process of real-time temperature control of the extraction kettle, the higher the parameter optimization factor corresponding to the real-time temperature monitoring time series data window of the extraction kettle, the higher the possibility of temperature fluctuations in the subsequent process, and the corresponding integral gain parameter is higher, thereby reducing the sensitivity of the control system and avoiding overshoot or oscillation due to temperature fluctuations in the temperature control process due to the sensitivity of the control system.

[0087] After obtaining the optimized parameter optimization factor, the temperature control of the supercritical extraction of plant essential oil is carried out according to the PID control model with the optimized parameters.

[0088] In summary, the embodiments of the present invention collect the temperature monitoring time series data in the supercritical extraction kettle of plant essential oil, obtain the historical temperature monitoring time series data of the extraction kettle, divide all the historical temperature monitoring time series data into windows to obtain all temperature time series windows; for any extraction kettle, perform clustering analysis according to the optimized distance metric between the historical temperature monitoring time series data windows to obtain the cluster division result. According to the cluster division result of the historical temperature monitoring time series data windows, obtain the hidden state of each historical temperature monitoring time series data window; according to the historical temperature monitoring time series data and the hidden state of the historical temperature monitoring time series data windows, establish a hidden Markov prediction model; predict through the hidden Markov prediction model according to the real-time temperature monitoring time series data of the extraction kettle to obtain the state transition probability matrix of the real-time temperature monitoring time series data of the extraction kettle; according to the state transition probability matrix and the state fluctuation evaluation of the clusters in the cluster division result, obtain the parameter optimization factor of the real-time temperature monitoring time series data of the extraction kettle; according to the historical temperature monitoring time series data of the extraction kettle, obtain a PID control model for the real-time temperature control of the extraction kettle, including the parameters after convergence of the PID control model; according to the parameter optimization factor of the real-time temperature monitoring time series data of the extraction kettle, optimize the parameters of the PID control model for the real-time temperature monitoring data of the extraction kettle, and perform temperature control for the supercritical extraction of plant essential oil according to the PID control model with optimized parameters. Among them, by performing fluctuation analysis on all the historical temperature monitoring time series data windows of the extraction kettle, an optimized distance metric between the historical temperature monitoring time series data windows is obtained, clustering is performed through this optimized distance metric to obtain a clustering result, each cluster is used as a temperature fluctuation state to determine the hidden state of each temperature time series data window, a hidden Markov prediction model is established through the temperature monitoring time series data and the hidden state of the temperature time series data windows, the real-time temperature monitoring time series data of the extraction kettle is used to obtain the state transition probability matrix through the hidden Markov prediction model, and the parameter optimization factor is obtained through the state transition probability matrix and the state fluctuation evaluation of each state, which is used to optimize the integral gain parameter of the PID model when regulating the real-time temperature of the extraction kettle through the PID model, so as to obtain more rigorous PID model parameters, thereby avoiding temperature overshoot or oscillation during the temperature regulation process and improving the control accuracy of the temperature control model.

[0089] The above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it; although the present invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that: they can still modify the technical solutions recorded in the foregoing embodiments, or perform equivalent replacements for some of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present invention, and should all be included in the protection scope of the present invention.

Claims

1. A temperature intelligent control method for supercritical extraction of plant essential oils, characterized in that: The method comprises: The temperature monitoring time series data in the supercritical extraction kettle of the plant essential oil is collected to obtain the historical temperature monitoring time series data of the extraction kettle, and all the historical temperature monitoring time series data are divided into windows to obtain all temperature time series windows; For any extraction kettle, cluster analysis is performed according to the optimized distance measurement between the historical temperature monitoring time series data windows to obtain clustering results; according to the clustering results of the historical temperature monitoring time series data windows, the hidden state of each historical temperature monitoring time series data window is obtained; A hidden Markov prediction model is established according to the historical temperature monitoring time series data and the hidden state of the historical temperature monitoring time series data window; a state transition probability matrix of the real-time temperature monitoring time series data of the extraction kettle is obtained by predicting the real-time temperature monitoring time series data of the extraction kettle through the hidden Markov prediction model; a parameter optimization factor of the real-time temperature monitoring time series data of the extraction kettle is obtained according to the state transition probability matrix and the state fluctuation evaluation of the clusters in the clustering result; According to the historical temperature monitoring time series data of the extraction kettle, a PID control model for real-time temperature control of the extraction kettle temperature is obtained, including the converged parameters of the PID control model; according to the parameter optimization factor of the real-time temperature monitoring time series data of the extraction kettle, the PID control model of the real-time temperature monitoring data of the extraction kettle is parameter optimized, and the temperature control of supercritical extraction of plant essential oil is performed according to the PID control model after the parameter optimization.

2. A temperature intelligent control method for supercritical extraction of plant essential oils according to claim 1, characterized in that: The cluster analysis is performed based on the optimized distance measurement between the historical temperature monitoring time series data windows to obtain the cluster classification results, including: For any historical temperature monitoring time series data window, obtain the timestamp distance between the data point and the center point of the window in the historical temperature monitoring time series data window, and use the inverse of the result of adding the timestamp distance to the constant 1 as the first influencing parameter, obtain the COF outlier detection of the data point in the historical temperature monitoring time series data window, and obtain the outlier factor of the data in the historical temperature monitoring time series data window. Normalize the absolute value of the outlier factor subtracted from the constant 1 as the second influencing parameter; multiply the first influencing parameter by the constant 1 and perform softmax normalization on the result of the second influencing parameter as the fluctuation analysis contribution degree corresponding to the data in the historical temperature time series data window; For any of the historical temperature monitoring time series data windows, according to the fluctuation analysis contribution of the data in the historical temperature monitoring time series data window, the temperature mean of the historical temperature monitoring time series data window and the historical temperature monitoring time series data window, the difference between the data in the historical temperature monitoring time series data window and the temperature mean of the historical temperature monitoring time series data window is used as the first fluctuation measure; according to the fluctuation analysis contribution corresponding to the data in the historical temperature time series data window, all the first fluctuation measures are weightedly summed and divided by the window length of the historical temperature time series data window to obtain the corresponding fluctuation calculation result; Obtain the slope in the least squares fitting result of the historical temperature time series data window, and use the absolute value of the difference between the fluctuation calculation result of the historical temperature monitoring time series data window and the fluctuation calculation result of the historical temperature monitoring time series data window corresponding to the cluster center in the clustering process as the first measure of the difference optimization factor; use the absolute value of the difference between the slope in the least squares fitting result of the historical temperature monitoring time series data window and the slope in the least squares fitting result of the historical temperature monitoring time series data window corresponding to the cluster center in the clustering process as the second measure of the difference optimization factor; normalize the product of the first measure of the difference optimization factor and the second measure of the difference optimization factor to obtain the corresponding normalized value as the difference optimization factor between the historical temperature monitoring time series data window and the historical temperature monitoring time series data window corresponding to the cluster center in the clustering process; For any of the historical temperature monitoring time series data windows in the clustering process, the Euclidean distance between the historical temperature monitoring time series data window and the historical temperature monitoring time series data window corresponding to the cluster center in the clustering process is used as the first Euclidean distance; the product of the difference optimization factor between the historical temperature monitoring time series data window and the historical temperature monitoring time series data window corresponding to the cluster center in the clustering process and the first Euclidean distance is used as the optimized distance metric between the historical temperature monitoring time series data window in the clustering process and the historical temperature monitoring time series data window corresponding to the cluster center in the clustering process; The clustering process is completed by using a K-means clustering algorithm based on the optimized distance measurement between the historical temperature monitoring time series data window and the historical temperature monitoring time series data window corresponding to the cluster center of the clustering process, and the clustering result of the historical temperature monitoring time series data window is obtained.

3. A temperature intelligent control method for supercritical extraction of plant essential oils according to claim 1, characterized in that: The step of obtaining the hidden state of each historical temperature monitoring time series data window according to the clustering result of the historical temperature monitoring time series data window includes: The clustering result of the historical temperature monitoring time series data window is obtained, and the cluster number of the historical temperature monitoring time series data window is used as the hidden state of each historical temperature monitoring time series data window.

4. A temperature intelligent control method for supercritical extraction of plant essential oils according to claim 1, characterized in that: The method of obtaining the state transition probability matrix of the real-time temperature monitoring time series data of the extraction kettle according to the hidden Markov prediction model of the historical temperature monitoring time series data includes: Obtain the hidden Markov model of the historical temperature monitoring time series data, and obtain the real-time temperature monitoring time series data of the extraction kettle; obtain the state transition probability matrix corresponding to the real-time temperature monitoring time series data of the extraction kettle according to the hidden Markov model of the historical temperature monitoring time series data.

5. A temperature intelligent control method for supercritical extraction of plant essential oils according to claim 2, characterized in that: The step of obtaining the parameter optimization factor of the real-time temperature monitoring time series data of the extraction kettle according to the state transition probability matrix and the state fluctuation evaluation of the clusters in the cluster division result includes: The state transition probability matrix of the real-time temperature monitoring time series data of the extraction kettle is obtained, and the state transition probability of the target state in the state transition probability matrix of the real-time temperature monitoring time series data of the extraction kettle is used as the fluctuation weight; the clustering result of the historical temperature monitoring time series data window of the clustering process is obtained, and the state fluctuation evaluation of each cluster in the clustering result is obtained according to the first fluctuation measure of the historical temperature monitoring time series data window in each cluster; the fluctuation weight of the real-time temperature monitoring time series data of the extraction kettle is weighted summed for the state fluctuation evaluations of all the clusters to obtain the corresponding weighted summation result; the opposite number of the weighted summation result is substituted into an exponential function with a natural constant e as the base, and the corresponding first exponential function result is obtained, and the subtraction result between the constant 1 and the first exponential function result is used as the parameter optimization factor of the real-time temperature monitoring time series data of the extraction kettle.

6. A temperature intelligent control method for supercritical extraction of plant essential oils according to claim 5, characterized in that: The obtaining of a state fluctuation assessment of each cluster in the cluster division result includes: The clustering result of the historical temperature monitoring time series data window is obtained, and the mean value of the fluctuation calculation of the historical temperature monitoring time series data window in each cluster in the clustering result of the historical temperature monitoring time series data window is used as the state fluctuation evaluation of the cluster.

7. The temperature intelligent control method for supercritical extraction of plant essential oil according to claim 1, characterized in that: The parameter optimization factor of the real-time temperature monitoring time series data of the extraction kettle is used to optimize the parameters of the PID control model of the real-time temperature monitoring data of the extraction kettle, including: The integral gain parameter of the real-time temperature PID control model of the extraction kettle and the parameter optimization factor of the real-time temperature monitoring time series data of the extraction kettle are respectively obtained; a constant 1 is added to the parameter optimization factor of the real-time temperature monitoring time series data of the extraction kettle as a first parameter weight, and the integral gain parameter of the real-time temperature PID control model of the extraction kettle is weighted according to the first parameter weight, and the result is used as the optimized integral gain parameter; the optimized integral gain parameter is used as the real-time integral gain parameter in the real-time temperature PID control model of the extraction kettle, and the PID control model after the parameter optimization is obtained.

8. The temperature intelligent control method for supercritical extraction of plant essential oils according to claim 1, characterized in that: The window division of all historical temperature monitoring time series data to obtain all temperature time series windows includes: The historical temperature monitoring time series data of the extraction kettle and the preset window length are respectively obtained; the historical temperature monitoring time series data of the extraction kettle are traversed continuously with a constant step length of 1 according to the preset window length, and the subsequence contained in each movement of the window of the preset window length is used as the historical temperature monitoring time series data window corresponding to the center point of the window.

Images