Control method of moisture content at the inlet of dried tofu based on multivariate time series analysis prediction model

Abstract

A method for controlling moisture content at the inlet of dried silk based on a multivariate time series analysis prediction model comprises: adopting a temperature and humidity control strategy suitable for large spaces in industrial plants to ensure dynamic stability of temperature and humidity control in the silk production area, collecting historical data related to the model, and normalizing the model input and output variables; using correlation analysis to extract features of input variables, and screening out the input variables most relevant to the moisture content at the inlet of dried silk as characteristic variables; establishing a multivariate time series analysis prediction model for moisture content at the inlet of dried silk according to the extracted characteristic variables; optimizing parameters of the multivariate time series analysis model; considering the time-varying characteristics of the model, training and updating the model in real time online; and reversely predicting the target moisture content value before entering the cabinet that makes the moisture content at the inlet of dried silk meet the standard according to the established model variable relationship and related characteristic data, so as to achieve the purpose of controlling the moisture content at the inlet of dried silk. The invention improves the accuracy of moisture content at the inlet of dried silk.

Description

Technical Field

[0001] The invention relates to the field of tobacco processing, and in particular to a method for controlling moisture content at a dried tobacco cut entrance based on a multivariate time series analysis prediction model. Background Art

[0002] The moisture content at the entrance of the drying wire is an important process indicator in the silk production process. It plays an important role in the process control and stability of the subsequent process. This indicator has received extensive attention from tobacco companies. There are several hours of leaf storage and shredding process before entering the cabinet to the entrance of the drying wire. Due to the influence of various factors, the intermediate links are prone to moisture loss, and there is no means to correct the moisture content of the material during the process. That is to say, the regulation of the moisture content at the entrance of the drying wire depends on the monitoring of the moisture content before entering the cabinet in the previous process.

[0003] The existing moisture content of the dried silk inlet is controlled by the operator manually estimating the target moisture content before entering the cabinet based on actual production, weather conditions, leaf storage time, etc. However, simple manual estimation cannot achieve the coordination and precise control of the parameters of the previous and next processes. Therefore, in order to ensure that the stability of the moisture content of the dried silk inlet during the silk making process meets the expectations, a more accurate and scientific moisture content control method is needed. Summary of the invention

[0004] The present invention provides a method for controlling the moisture content of dried wire at an inlet based on a multivariate time series analysis prediction model. The moisture content of dried wire at an inlet is predicted and controlled by adjusting a target value of the moisture content before entering a cabinet, so that a batch mean value of the moisture content of dried wire at an inlet reaches a process target value. The target moisture value before entering a cabinet is automatically pushed, replacing a manual estimation method, thereby improving the accuracy of the moisture content of dried wire at an inlet.

[0005] According to an embodiment, a method for controlling moisture content at a dried tofu inlet based on a multivariate time series analysis prediction model is provided, comprising the following steps:

[0006] Control the ambient temperature and humidity in the silk production area, and obtain historical data related to the multivariate time series analysis prediction model under the premise of stable ambient temperature and humidity control. The historical data includes moisture before entering the cabinet, silk drying ambient temperature, silk drying ambient humidity, silk drying inlet moisture and leaf storage time, and perform normalization on the input and output variables related to the multivariate time series analysis prediction model;

[0007] Correlation analysis was used to extract the features of input variables, and the input variables most relevant to the moisture content of dried silk inlet were selected as feature variables.

[0008] According to the extracted characteristic variables, a multivariate time series analysis and prediction model of moisture content at the inlet of dried tofu was established based on the data-driven method.

[0009] According to the graphical order determination method and Bayesian information criterion, the parameters of the multivariate time series analysis and prediction model are optimized;

[0010] Considering the time-varying characteristics of the model, the sliding window algorithm is used to train and update the multivariate time series analysis and prediction model in real time online;

[0011] According to the established multivariate time series analysis prediction model variable relationship and related characteristic data, the target moisture content before entering the cabinet that can make the moisture content of the dried strands meet the standard is reversely predicted, thus realizing the control of the moisture content of the dried strands at the inlet.

[0012] In an embodiment, the method for normalizing the input and output variables related to the multivariate time series analysis prediction model is:

[0013]

[0014] Among them, x0 and x i They represent the variables before and after normalization, Maxx represents the maximum value of the x variable, and Minx represents the minimum value of the x variable.

[0015] In the embodiment, the method for extracting the features of the input variables according to the correlation analysis is: the variables most correlated with the moisture content of the dried tofu inlet are screened out according to the Pearson correlation coefficient between the variables, and the Pearson correlation coefficient is expressed as:

[0016]

[0017] In the formula, r is the Pearson correlation coefficient, X i With Y i are represented as the i-th sample in the X and Y variables, respectively. and are the means of X and Y respectively. After obtaining the correlation coefficient of each input variable to the moisture content of the dried tofu inlet, several variables with the largest correlation coefficients are selected as the input variables of the model to complete the feature extraction.

[0018] In the embodiment, the established multivariate time series analysis prediction model of the moisture content of the dried tofu inlet is expressed as:

[0019]

[0020] In the formula, y t is the moisture content of the dried silk at the entrance at time t, y t-i is the moisture content of the dried silk at the historical moment ti, Φ i is the coefficient of the ith autoregressive term, p is the order of the autoregressive term, ε t-j is the forecast error of the model at the historical moment tj, q is the order of the sliding average term, θ j is the jth sliding average coefficient, Xz,t is the Zth input variable at time t, m is the number of input variables, including moisture before entering the cabinet, drying environment temperature, drying environment humidity and leaf storage time, etc., β Z is the model coefficient corresponding to the Zth input variable, ε t is the model prediction error at time t.

[0021] In the embodiment, the method for optimizing parameters of the multivariate time series analysis prediction model is: using the method of determining the order by looking at the picture to select the most suitable stable minimum difference order d, by observing the time series diagram of the predicted data, selecting the first-order or second-order difference, if the time series of the predicted data itself is stable, then no difference term is required, and then using the Bayesian information criterion to select the autoregressive term order p and the sliding average term order q, the BIC criterion is expressed as:

[0022] BIC=ln(n)k-2ln(L)

[0023] Where k is the number of model parameters, n is the number of samples, and L is the likelihood function.

[0024] In an embodiment, the method of using the sliding window algorithm to train and update the model in real time online is: every time a new batch of samples is obtained, the earliest batch of samples in the model training samples are eliminated to form new training samples for model training.

[0025] In an embodiment, a method for controlling the environmental temperature and humidity of a silk production area includes: setting the central values ​​of the monthly temperature and relative humidity of the silk production area according to the historical data of the outdoor temperature and humidity of each month throughout the year; formulating the control range of the temperature and relative humidity of the large space in the silk production area according to the capacity of the air-conditioning system and in combination with the production and manufacturing process requirements; collecting temperature and humidity data through temperature and humidity sensor nodes arranged in the silk production area, and sending them to a controller in real time; the controller compares the temperature and humidity data of the silk production area with the central values ​​of the temperature and humidity of the corresponding month in real time through a PID automatic control mode, and sends heating, cooling, humidification, and dehumidification instructions to the temperature and humidity adjustment equipment according to the results, and performs corresponding work to keep the fluctuations of the temperature and relative humidity of the large space in the silk production area within the control range.

[0026] At least one embodiment of the present invention provides a tow inlet moisture content control system based on a multivariate time series analysis prediction model, comprising: a processor; a memory for storing processor executable instructions; wherein the processor is configured to complete all or part of the steps of the method.

[0027] At least one embodiment of the present invention provides a non-transitory computer-readable storage medium having a computer program stored thereon. When the computer program is executed by a processor, all or part of the steps of the method are completed.

[0028] The beneficial effect of the present invention is that, by combining historical data and the current production process mechanism, a multivariate time series prediction model is established, and the moisture content at the entrance of the dried silk is predicted according to the key related variables of the storage leaf shredding process, and the target moisture content before entering the cabinet that can make the moisture content at the entrance of the dried silk meet the standard is inferred in advance. Compared with the traditional method, the multivariate time series model takes into account the time series characteristics of the key variables themselves, which can not only unify the adjustment method of the target moisture content before entering the cabinet, reduce the human system error caused by different operation methods, but also reduce the influence of hysteresis on moisture content control, and realize the smooth control of moisture content at the entrance of the dried silk. BRIEF DESCRIPTION OF THE DRAWINGS

[0029] In order to more clearly illustrate the technical solution of the embodiment of the present invention, the drawings of the embodiment are briefly introduced below.

[0030] Figure 1 A schematic flow chart of a method for controlling moisture content at a drying tofu inlet based on a multivariate time series analysis and prediction model provided in one embodiment of the present invention.

[0031] Figure 2 A schematic diagram of the relationship between variables in a multivariate time series model of moisture content at the inlet of dried tofu provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0032] like Figure 1 , Figure 2 The present invention provides a method for controlling the moisture content of dried wire inlet based on a multivariate time series analysis prediction model. The moisture content of dried wire inlet is predicted and controlled by adjusting the moisture content target value before entering the cabinet, so that the batch mean of moisture content of dried wire inlet reaches the process target value, and the moisture target value before entering the cabinet is automatically pushed, replacing the manual estimation method, and improving the compliance and accuracy of moisture content of dried wire inlet. Figure 1 , Figure 2 The present invention is described in detail.

[0033] First, the ambient temperature and humidity of the silk production area are controlled within the set range. A temperature and humidity control strategy suitable for the tall and large spaces of industrial plants is adopted. By arranging corresponding temperature and humidity sensor nodes in the tobacco silk production area, the actual outdoor temperature and humidity differences in each month of the year are considered under the premise of the tall and large space of the industrial plant, and the monthly indoor temperature and humidity differentiation center value is set to ensure the dynamic stability of the temperature and relative humidity control in the silk production area.

[0034] (1) Based on the historical data of outdoor temperature and humidity for each month of the year, formulate the annual temperature and humidity control standards, that is, the central values ​​of temperature and relative humidity for each month;

[0035] (2) According to the equipment capacity of the industrial plant air conditioning system and the production and manufacturing process requirements, the control range of temperature and relative humidity in the large space of the industrial plant is formulated, such as the control range of ambient temperature and humidity within ±2°C and ±2%;

[0036] (3) Indoor temperature and humidity sensors collect temperature and humidity data in the production area and send them to the controller in real time;

[0037] (4) The controller uses the PID automatic control mode to compare the temperature and humidity data of the production area with the corresponding monthly temperature and humidity center values ​​in real time, and sends heating, cooling, humidification, and dehumidification instructions to the temperature and humidity adjustment equipment based on the results to perform corresponding tasks, so that the fluctuations of temperature and relative humidity in the large space of the industrial plant are within the control range.

[0038] Table 1 Temperature and humidity control plan for the silk-making area throughout the year

[0039] Serial number start date End Date Monthly duration Temperature center value ℃ Humidity center value % Temperature and humidity tolerance 1 March 1 April 30 2 26 55 ±2 2 May 1 June 30 2 26 58 ±2 3 July 1 September 30 3 28 58 ±2 4 October 1 November 30 2 26 55 ±2 5 December 1 February 28 3 24 55 ±2

[0040] Under the premise of stable control of the above-mentioned environmental temperature and humidity, a method for controlling the moisture content at the drying inlet based on a multivariate time series analysis prediction model is implemented.

[0041] Step 1: Collect historical data related to the model, including moisture before entering the cabinet, drying environment temperature, drying environment humidity, drying inlet moisture and leaf storage time, etc., normalize the model input and output variables, and convert variables of different dimensions into values ​​between 0 and 1 to reduce the impact of dimensional error introduced by the unit range of the dependent variable itself.

[0042]

[0043] Among them, x0 and x i They represent the variables before and after normalization, Maxx represents the maximum value of the x variable, and Minx represents the minimum value of the x variable.

[0044] Step 2, use correlation analysis to extract the features of input variables, and select the input variables most relevant to the moisture content of the drying silk inlet as feature variables, including moisture before entering the cabinet, drying silk environment temperature, drying silk environment humidity, leaf storage environment temperature, leaf storage environment humidity and leaf storage time, etc.

[0045] The variables most correlated with the moisture content at the entrance of dried tofu are screened out based on the Pearson correlation coefficient between the variables. The Pearson correlation coefficient is expressed as:

[0046]

[0047] In the formula, r is the Pearson correlation coefficient, X i With Y i Represented as the i-th sample in X and Y variables, respectively. and are the means of X and Y, respectively. After obtaining the correlation coefficient of each input variable to the moisture content of the dried tofu inlet, several variables with the largest correlation coefficients are selected as the input variables of the model to complete the feature extraction.

[0048] Step 3: Based on the extracted characteristic variables, a multivariate time series analysis and prediction model of the moisture content of the dried tofu inlet is established based on the data-driven method. The model can be expressed as:

[0049]

[0050] In the formula, y t is the moisture content of the dried silk at the entrance at time t; t-i is the moisture content of the dried silk at the entrance at the historical moment ti; Φ i is the coefficient of the ith autoregressive term; p is the order of the autoregressive term; ε t-j is the prediction error of the model at the historical moment tj; q is the order of the sliding average term; θ j is the coefficient of the jth sliding average term; X z,t is the Zth input variable at time t; m is the number of input variables, including moisture before entering the cabinet, drying environment temperature, drying environment humidity and leaf storage time, etc.; β Z is the model coefficient corresponding to the Zth input variable, ε t is the model prediction error at time t.

[0051] Step 4: Optimize the parameters of the multivariate time series analysis model based on the order determination method and the Bayesian Information Criterion (BIC) to improve the model prediction performance. The most suitable stable minimum difference order d is selected by the order determination method. By observing the time series diagram of the predicted data, the first or second order difference can be selected. If the time series of the predicted data is stable, then no difference term is required. Then the Bayesian Information Criterion is used to select the autoregressive term order p and the sliding average term order q. The BIC criterion can be expressed as:

[0052] BIC=ln(n)k-2ln(L)

[0053] Among them, k is the number of model parameters, n is the number of samples, and L is the likelihood function. The smaller the BIC value, the better the model parameters, and thus the better the model performance.

[0054] Step 5, considering the time-varying characteristics of the model, the sliding window algorithm is used to train and update the model in real time online, that is, the model training sample capacity is fixed, and whenever the data acquisition system obtains a new batch of samples, the earliest batch of samples in the model training samples are removed to form new training samples for model training, thereby updating the model online. This ensures the self-correction and self-learning properties of the model, and further optimizes the accuracy of the control system.

[0055] Step 6, based on the established model variable relationship and related characteristic data, reversely predict the target moisture content before entering the cabinet that can make the moisture content of the dried tofu inlet meet the standard, so as to achieve the purpose of controlling the moisture content of the dried tofu inlet.

[0056] In an exemplary embodiment, a tow inlet moisture content control system based on a multivariate time series analysis prediction model is also provided, comprising: a memory; a memory for storing processor executable instructions; a processor, wherein the processor is configured to execute the instructions in the memory to complete all or part of the steps of the above-mentioned method.

[0057] In an exemplary embodiment, a non-transitory computer-readable storage medium including instructions is also provided, such as a memory including instructions, and the instructions can be executed by a processor to complete all or part of the steps of the above method. For example, the non-transitory computer-readable storage medium can be a ROM, RAM, CD-ROM, magnetic tape, floppy disk, optical data storage device, etc.

Claims

1. A method for controlling moisture content of dried tofu inlet based on a multivariate time series analysis prediction model, characterized in that: include: Control the ambient temperature and humidity in the silk production area, and obtain historical data related to the multivariate time series analysis prediction model under the premise of stable ambient temperature and humidity control. The historical data includes moisture before entering the cabinet, silk drying ambient temperature, silk drying ambient humidity, silk drying inlet moisture and leaf storage time, and perform normalization on the input and output variables related to the multivariate time series analysis prediction model; Correlation analysis is used to extract the features of input variables, and the input variables most relevant to the moisture content of the dried silk inlet are selected as feature variables. The variables most relevant to the moisture content of the dried silk inlet are selected based on the Pearson correlation coefficient between the variables. The Pearson correlation coefficient is expressed as: In the formula, r is the Pearson correlation coefficient, X i With Y i are represented as the i-th sample in the X and Y variables, respectively. and are the means of X and Y respectively. After obtaining the correlation coefficient of each input variable to the moisture content of the dried tofu inlet, several variables with the largest correlation coefficients are selected as the input variables of the multivariate time series analysis and prediction model. According to the extracted characteristic variables, a multivariate time series analysis and prediction model of moisture content at the inlet of dried tofu is established based on the data-driven method: In the formula, y t is the moisture content of the dried silk at the entrance at time t, y t-i is the moisture content of the dried silk at the historical moment ti, Φ i is the coefficient of the ith autoregressive term, p is the order of the autoregressive term, ε t-j is the forecast error of the model at the historical moment tj, q is the order of the sliding average term, θ j is the jth sliding average coefficient, X z,t is the Zth input variable at time t, m is the number of input variables, including moisture before entering the cabinet, drying environment temperature, drying environment humidity and leaf storage time, β Z is the model coefficient corresponding to the Zth input variable, ε t is the model prediction error at time t; According to the method of determining the order by looking at the graph and the Bayesian information criterion, the parameters of the multivariate time series analysis and prediction model are optimized: the method of determining the order by looking at the graph is used to select the most suitable stable minimum difference order d, and by observing the time series diagram of the predicted data, the first or second order difference is selected. If the time series of the predicted data is stable, the difference term is not required, and then the Bayesian information criterion is used to select the autoregressive term order p and the sliding average term order q. The BIC criterion is expressed as: BIC=ln(n)k-2ln(L) In the formula, k is the number of model parameters, n is the number of samples, and L is the likelihood function; Considering the time-varying characteristics of the model, the sliding window algorithm is used to train and update the multivariate time series analysis and prediction model in real time online; According to the established multivariate time series analysis prediction model variable relationship and related characteristic data, the target moisture content before entering the cabinet that can make the moisture content of the dried strands meet the standard is reversely predicted, thus realizing the control of the moisture content of the dried strands at the inlet.

2. The method according to claim 1, characterized in that The method for normalizing the input and output variables related to the multivariate time series analysis and prediction model is: Among them, x0 and x i They represent the variables before and after normalization, Maxx represents the maximum value of the x variable, and Minx represents the minimum value of the x variable.

3. The method according to claim 1, characterized in that The method of using the sliding window algorithm to train and update the model in real time online is: every time a new batch of samples is obtained, the earliest batch of samples in the model training samples are eliminated to form new training samples for model training.

4. The method according to claim 1, characterized in that Methods for controlling the ambient temperature and humidity in the silk production area include: According to the historical data of outdoor temperature and humidity in each month of the year, the central values ​​of temperature and relative humidity in the monthly silk production area are set; According to the capacity of the air conditioning system and the requirements of the manufacturing process, the control range of temperature and relative humidity in the large space of the silk production area is formulated; The temperature and humidity sensor nodes arranged in the silk production area collect temperature and humidity data and send them to the controller in real time; Through the PID automatic control mode, the controller compares the temperature and humidity data of the silk production area with the temperature and humidity center values ​​of the corresponding month in real time, and sends heating, cooling, humidification and dehumidification instructions to the temperature and humidity adjustment equipment according to the results to perform corresponding work, so that the fluctuations of temperature and relative humidity in the large space of the silk production area are within the control range.

5. A moisture content control system for dried tofu inlet based on a multivariate time series analysis prediction model, characterized in that: include: processor; A memory for storing processor-executable instructions; wherein the processor is configured to execute the steps of the method according to any one of claims 1 to 3.

6. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 3 are implemented.

Images