KR20220046158A - Analysis method of mixed ratio of waste cooking-oils for biodiesel production process - Google Patents

Abstract

The present invention relates to an analysis method which develops a PLSR model to non-destructively measure a raw material mixing ratio of waste cooking oil for predicting an amount of each raw material in mixed waste cooking oil by using a near-infrared spectroscope. The present invention can determine the PLSA model indicating high linearity and a low prediction error and can greatly contribute to improvement of biodiesel yield when applying the PLSA model to a biodiesel production process using the waste cooking oil.

Description

Analysis method of mixed ratio of waste cooking-oils for biodiesel production process applicable to biodiesel production process

The present invention relates to a method for analyzing the raw material mixing ratio of waste cooking oil, and more particularly, to an analysis method for non-destructively measuring the mixing ratio of the waste cooking oil raw material mixed in a biodiesel production process.

Background to the present disclosure is provided herein, which does not necessarily imply known art.

Recently, as interest in environmental problems has rapidly increased and fossil fuels have been increasingly depleted, biodiesel has received a lot of attention as an alternative fuel to fossil fuels, and accordingly, many studies on biodiesel are being conducted. Compared to general diesel fuel, biodiesel emits less greenhouse gases, such as carbon dioxide, sulfur, and suspended particles, as well as substances that cause air pollution. In addition, it has the advantage of preserving the amount of carbon in the entire planet without increasing the amount of carbon that accelerates global warming.

As raw materials for biodiesel production, algae, palm oil, coconut oil, and various types of cooking oil are mainly used internationally. However, since these raw materials are expensive, the production cost incurred when these raw materials are used as raw materials for biodiesel production is more expensive than petroleum diesel production cost, so it is economically difficult to mass-produce and develop biodiesel.

It is economically and environmentally advantageous to replace the existing raw materials used in the production of biodiesel with waste cooking oil. Since the price of waste cooking oil is about half that of virgin oil, if waste cooking oil is used as a raw material for biodiesel production, the production cost of biodiesel can be greatly reduced. In addition, if waste cooking oil, which was previously difficult to dispose and dispose of, is used for biodiesel production, it can reduce water pollution and help the drainage environment.

Although the amount of waste cooking oil generated varies from country to country depending on the amount of cooking oil used, a very large amount of waste cooking oil is discharged every year. According to an ECOFYS survey, as of 2013, European countries excluding the UK produced 1 million tons of waste cooking oil, and non-European countries (USA, China, Indonesia and Argentina) produced about 4.5 million tons of waste cooking oil. And as the world's population increases, it is estimated that the emission of waste cooking oil will increase further in the future.

The production of biodiesel using waste cooking oil is generally made through a transesterification reaction in which methyl ester and glycerol are produced by adding alcohol and a catalyst to fats and oils. In the transesterification reaction, an alkali catalyst or an acid catalyst is widely used as a catalyst for accelerating the rate of the main reaction, and methanol is usually used as an alcohol. Through transesterification, free fatty acids contained in triglycerides, the main component of waste cooking oil, are converted into methyl esters, which are then used as biodiesel.

The edible oil raw materials used in the biodiesel production process have different types of fatty acids and different proportions. However, even if these raw materials are heated and turned into waste cooking oil, the proportion of fatty acids constituting each raw material does not change significantly.

The type of fatty acid constituting the raw material of the waste cooking oil and the ratio of the raw material to the raw material affect variables such as the amount of alcohol required in the biodiesel production process using the waste cooking oil, the type of catalyst, the amount of the catalyst, and the reaction temperature, and eventually the biodiesel production process. Affects diesel yield. That is, when the raw material mixing ratio of the waste cooking oil is changed, the type and composition ratio of fatty acids constituting the mixed waste cooking oil changes. Therefore, the optimal amount of alcohol required for the transesterification reaction in the biodiesel production process, the type and amount of the catalyst, and the reaction temperature is decided Therefore, it is very important to know the raw material mixing ratio of waste cooking oil first in selecting the parameters necessary for the biodiesel production process.

However, although there have been studies on the classification of raw materials for cooking oil, it is difficult to find studies on the analysis of the raw material mixing ratio of waste cooking oil.

Differential thermal analysis (Differential Thermal Analyzer), High-Performance Liquid Chromatography (HPLC) analysis, Gas Chromatography (GC) analysis, etc. are used for quantitative analysis of oils and fats. These assays are highly accurate, but suffer from sensitivity or limitations, and are not continuous and destructive.

The present invention provides an analysis system for non-destructively measuring the raw material mixing ratio of waste cooking oil by developing a PLSR model to predict the amount of each raw material in the mixed waste cooking oil using a near-infrared spectrometer.

However, the objects of the present invention are not limited to the above-mentioned objects, and other objects not mentioned will be clearly understood by those skilled in the art from the following description.

The present invention is a data collection step (S1) of mixing at least two or more kinds of waste cooking oil to collect near-infrared absorption spectrum data for each concentration of the mixed raw material; a modeling step (S2) of obtaining a prediction candidate model according to a wavelength region using a multivariate statistical analysis (Partial Least Squares Analysis; PLSA) method from the obtained near-infrared absorption spectrum data; With respect to the obtained prediction candidate model, the prediction candidate model of the wavelength region having the highest prediction performance is predicted using any one or more of a coefficient of determination (R ² ) and a root mean square error (RMSE). a model determination step (S3) of determining a model; and an analysis step (S4) of predicting the content of a specific raw material by measuring the near-infrared absorption spectrum of the mixed waste cooking oil sample to be analyzed and then substituting it into the prediction model of the determined wavelength region; provide a way

In addition, in the modeling step (S2), the wavelength region is a first region from 898.77 nm to 2132.65 nm, a second region from 1,100 nm to 1250 nm, a third region from 1,600 nm to 1800 nm, and 1,100 nm to 1250 nm and 1,600 nm to 1800 nm. It is characterized in that it includes a fourth region.

In addition, the model determining step ( S3 ) is a step of determining, as a prediction model, a prediction candidate model in the wavelength region having the highest prediction performance compared with an actual value among the prediction candidate models according to the wavelength region.

The model determination step (S3) for predicting the content of the waste sunflower oil when the waste cooking oil is a mixture of sunflower oil and waste canola oil is a prediction candidate in the first region from 898.77 nm to 2132.65 nm It is characterized in that it is a step of determining the model as a predictive model.

The model determination step (S3) for predicting the content of waste olive oil when the waste cooking oil is a mixture of waste soybean oil, waste olive oil and waste sunflower oil (S3) is 1600nm to 1800nm It is characterized in that it is a step of determining a prediction candidate model of three regions as a prediction model.

When the waste cooking oil is mixed with waste soybean oil, waste olive oil, waste canola oil and waste sunflower oil, the model determination step (S3) for predicting the content of waste olive oil ) is a step of determining a prediction candidate model of the first region ranging from 898.77 nm to 2132.65 nm as a prediction model.

The present invention can determine a PLSA model showing high linearity and low prediction error by generating a PLS model according to the wavelength region through the absorption spectrum obtained using a near-infrared spectrometer, and through this PLSA model, a specific The content of raw materials can be predicted non-destructively. Accordingly, if the near-infrared spectroscopy and PLSA model are applied to the biodiesel production process using waste edible oil, it provides an effect that can greatly contribute to the improvement of biodiesel yield.

1 shows an image of a near-infrared region absorbance measurement of a waste cooking oil sample according to an embodiment of the present invention.
Figure 2 shows the PCA modeling results of waste cooking oil and reproductive oil according to an embodiment of the present invention.
3 shows the absorption spectrum data of 8 types of mixing waste soybean oil and waste olive oil according to an embodiment of the present invention, and changing the concentration of waste soybean oil from 0.3 ml to 2.4 ml.
4 is a PLSR model for predicting the amount of waste olive oil in the waste oil mixture (OS) in a static state according to an embodiment of the present invention (A) and PLSR predicting the amount of waste olive oil in the waste oil mixture (OS) in a static state according to an embodiment of the present invention; The model verification result (B) is shown.

All technical and scientific terms used herein have the same meaning as commonly understood by those of ordinary skill in the art, unless otherwise stated. Also throughout this specification and claims, unless otherwise indicated, the term comprise, comprises, comprising is meant to include the recited object, step or group of objects, and steps, and any other It is not used in the sense of excluding an object, step, or group of objects or groups of steps.

Prior to describing the present invention in detail below, it is to be understood that the terminology used herein is for the purpose of describing specific embodiments and is not intended to limit the scope of the present invention, which is limited only by the appended claims. shall.

On the other hand, various embodiments of the present invention may be combined with any other embodiments unless clearly indicated to the contrary. Any feature indicated as particularly preferred or advantageous may be combined with any other feature and features indicated as preferred or advantageous. Hereinafter, embodiments of the present invention and effects thereof will be described with reference to the accompanying drawings.

The raw material mixing ratio analysis method of the waste cooking oil according to an embodiment of the present invention comprises the steps of: collecting near-infrared absorption spectrum data for each concentration of each of the mixed raw materials by mixing at least two or more kinds of waste cooking oil (S1); a modeling step (S2) of obtaining a prediction candidate model according to a wavelength region using a multivariate statistical analysis (Partial Least Squares Analysis; PLSA) method from the obtained near-infrared absorption spectrum data; Determining a prediction model of a wavelength region having the highest prediction performance by using any one or more of a coefficient of determination (R ² ) and a root mean square error (RMSE) with respect to the obtained prediction candidate model model determination step (S3); and an analysis step (S4) of predicting the content of a specific raw material by measuring the near-infrared absorption spectrum of the mixed waste cooking oil sample to be analyzed and then substituting it into the prediction model of the determined wavelength region.

The data collection step (S1) is a step of mixing at least two or more kinds of waste cooking oil to collect near-infrared absorption spectrum data for each concentration of each mixed raw material, and analysis data for determining a prediction model and predicting a mixing ratio of a specific raw material is the step of collecting

The data collection step (S1) measures the near-infrared absorption spectrum according to the mixing type and mixing ratio of the waste cooking oil. For example, when two types of waste cooking oil are mixed by n concentrations, a total of 2n near-infrared absorption spectral data are collected, and when any two types of waste cooking oil among 4 types of waste cooking oil are mixed by n concentrations, the total Collect 6n near-infrared absorption spectral data.

The modeling step (S2) is a step of obtaining a prediction candidate model according to a wavelength region using a Partial Least Squares Analysis (PLSA) method from the obtained near-infrared absorption spectrum data. At least two wavelength regions are classified. Then, a prediction candidate model is obtained according to each classified wavelength region. Preferably, it is classified into four wavelength regions to obtain four prediction candidate models for each type of mixture and a specific raw material.

The four wavelength regions include a first region from 898.677 nm to 2132.65 nm, a second region from 1,100 nm to 1250 nm, a third region from 1,600 nm to 1800 nm, and a fourth region from 1,100 nm to 1250 nm and 1,600 nm to 1800 nm . The first region is the entire wavelength band, the second region is the second overtone region of the C-H bond not including the absorbance wavelength of water, the third region is the first harmonic region of the C-H bond, and the fourth region is the second region and the third region It is a combined area.

For example, when two types of waste cooking oil are mixed, a total of 8 (=2x4) prediction candidate models are obtained for each of the 4 wavelength ranges for each of the 2 types of mixed raw materials, and any 2 of the 4 types of waste cooking oil are obtained. In the case of mixing waste cooking oil of different species, there are 6 mixing cases, and a total of 48 (=6x2x4) prediction candidate models are obtained for each of the two mixed raw materials for each of the 4 wavelength regions.

The model determining step S3 uses any one or more of a coefficient of determination (R ² ) and a root mean square error (RMSE) for the obtained prediction candidate model to have the highest prediction performance. As a step of determining a prediction model, it is a step of determining a prediction model of a wavelength range having the highest prediction performance compared with an actual value among prediction candidate models according to a wavelength range obtained for each type of mixture of the spent cooking oil and a specific raw material.

The model determining step (S3) is the difference between the coefficient of determination (R ² ) and the actual value and the value predicted by the model for each prediction candidate model according to the type of mixture of the waste cooking oil and the wavelength range for each specific raw material. It is determined by comparing the root mean square error (RMSE). That is, among the prediction candidate models according to at least two wavelength regions, the prediction candidate model of the wavelength region having the closest coefficient of determination to 1 and the smallest root mean square error is determined as the prediction model for the corresponding mixture type and specific raw material.

For example, as shown in an experimental example to be described later, when waste olive oil (O) and waste sunflower oil (S) are mixed, R ² and By comparing the RMSE, the prediction candidate model of the first region, which has the highest R ² and the lowest RMSE, is determined as the prediction model with the best performance.

The analysis step (S4) is a step of predicting the content of a specific raw material by measuring the near-infrared absorption spectrum of the mixed waste cooking oil sample to be analyzed and then substituting it into the determined prediction model. The mixing ratio is predicted by substituting the measurement results into the prediction model.

The mixing ratio prediction model of the specific raw material of the mixed waste cooking oil determined according to the present invention shows high linearity and low prediction error, and the content of the specific raw material from the absorption spectrum of the mixed waste cooking oil through the generated PLSA model is non-destructively predictable. Accordingly, the application of near-infrared spectroscopy and PLS models to the biodiesel production process using waste edible oil can greatly contribute to the improvement of biodiesel yield.

<Examples and Experimental Examples>

(One) Waste Cooking Oil Sample Preparation

There are four types of waste cooking oil: waste soybean oil (B), waste sunflower oil (S), waste canola oil (C), and waste olive oil (O) that can be easily obtained in the market was prepared. Waste soybean oil, waste sunflower oil, and waste canola oil were obtained from a nearby restaurant. Waste olive oil was obtained by purchasing raw cooking oil and heating it for a long time in a laboratory after adding frozen food.

To measure the near-infrared (NIR) absorbance of the mixed waste cooking oil, two to four of the four raw materials were selected, and then each raw material was differently combined for each concentration so that the total volume was 3 ml. The concentration of each raw material was 0.3ml to 2.4ml, and samples were prepared by mixing 8 kinds of concentrations.

(2) Measurement of NIR Absorption Spectrum of Mixed Waste Cooking Oil Sample

The NIR absorbance of the waste cooking oil sample was measured with a NIR spectrometer (NIRQUEST, Ocean Optics, USA) (c) with the sample placed in a cuvette (b) with a 12 mm transmission distance and irradiated with a tungsten halogen light source (a) as shown in FIG. The wavelength band from 898.677 nm to 2132.65 nm was measured three times using the As a reference, the spectrum of an empty cuvette was used.

(3) NIR absorbance spectrum analysis

The obtained absorbance spectrum was used to develop a PLSR (partial least squares regression) model that can predict the amount of raw material in the mixed waste cooking oil sample through Unscrambler (ver. 9.7, CAMO Software, Norway).

The wavelength ranges of the absorbance data obtained from NIR spectroscopy were 1,100 nm to 1250 nm (first region), 1,600 nm to 1800 nm (second region), 1,100 nm to 1250 nm, and 1,600 nm to 1800 nm (third region), 898.677 nm to 2132.65 The nm (fourth region) was divided into four types and used to develop the PLSR model.

The cross validation method was used as a validation method for the predictability of the model.

(4) Analysis result - 1

First, the absorbance of four unmixed types of waste cooking oil (B', C', O', S') and raw food oil (B, C, O, S) was measured. Based on the acquired absorbance spectrum data, a Principle Component Analysis (PCA) model was developed. The developed PCA model showed the absorbance distribution of raw cooking oil on the left and waste cooking oil on the right based on a straight line passing through the origin as shown in FIG.

(5) Analysis result - 2

In FIG. 3, the absorption spectrum data of 8 types of waste soybean oil and waste olive oil were mixed and the concentration of the waste soybean oil was changed from 0.3 ml to 2.4 ml. As shown in FIG. ³ , the absorbance spectrum of the mixed waste cooking oil showed a similar shape without visible significant change depending on the concentration of the raw material in the mixed waste cooking oil. It was possible to distinguish the concentration through

(6) Analysis result - 3

The absorbance of the obtained blended waste cooking oil was developed to predict the amount of specific raw materials in the blended waste cooking oil sample. The absorbance data used to develop the multivariate statistical analysis (Partial Least Squares-Discriminant Analysis) model was classified into a total of four types as the region of the NIR wavelength band representing the composition ratio of fatty acids constituting the waste cooking oil.

The absorbance wavelength band selected from the first type is from 898.677 nm to 2132.65 nm, which is the entire wavelength band, the second type is from 1,100 nm to 1250 nm, the second overtone region of the C-H bond that does not include the absorbance wavelength of water, and the third type is C-H The first overtone region of coupling is 1,600 nm to 1800 nm, and finally, the fourth type is 1,100 nm to 1250 nm and 1,600 nm to 1800 nm, the second and third types combined. Based on this, a total of 12 models were created for each type.

The generated PLSR models consist of 12 models that predict the amount of a specific raw material in a sample mixed with two types of waste cooking oil. The performance of the model was compared using R ² , which is the correlation coefficient, and RMSE (Root Mean Square Error), which is the difference between the actual value and the value predicted by the model.

The results of the PLSR model predicting the amount of each waste cooking oil (WCO) component are shown in Tables 1 and 2 below. As shown in Tables 1 and 2 below, the PLSR model of the first type (first region) generated based on the absorbance spectrum of the entire wavelength showed the highest performance in predicting the amount of specific raw materials in the mixed waste cooking oil.

WCO
Mixture WCO Component Wavelength [nm] R ² _cal R ² _val RMSEC RMSEV decision
Model B-C B 1100-1250 0.998 0.997 0.113 0.148 1600-1800 0.998 0.997 0.109 0.133 1100-1250, 1600-1800 0.998 0.997 0.113 0.148 full spectrum 0.998 0.997 0.094 0.131 O C 1100-1250 0.993 0.974 0.201 0.395 1600-1800 0.998 0.997 0.109 0.135 1100-1250, 1600-1800 0.998 0.996 0.113 0.143 full spectrum 0.998 0.997 0.094 0.134 O B-O B 1100-1250 0.997 0.996 0.141 0.172 O 1600-1800 0.987 0.985 0.288 0.314 1100-1250, 1600-1800 0.991 0.988 0.256 0.282 full spectrum 0.990 0.990 0.247 0.272 O 1100-1250 0.997 0.996 0.141 0.171 O 1600-1800 0.987 0.983 0.288 0.324 1100-1250, 1600-1800 0.991 0.989 0.236 0.255 full spectrum 0.991 0.991 0.247 0.261 B-S B 1100-1250 0.991 0.986 0.234 0.307 1600-1800 0.996 0.994 0.150 0.223 1100-1250, 1600-1800 0.997 0.995 0.132 0.180 full spectrum 0.997 0.995 0.130 0.171 O S 1100-1250 0.991 0.985 0.234 0.311 1600-1800 0.996 0.991 0.150 0.224 1100-1250, 1600-1800 0.997 0.995 0.132 0.178 full spectrum 0.997 0.996 0.130 0.175

WCO
Mixture WCO Component Wavelength [nm] R ² _cal R ² _val RMSEC RMSEV decision
Model C-O C 1100-1250 0.989 0.978 0.262 0.385 1600-1800 0.991 0.989 0.242 0.297 1100-1250, 1600-1800 0.992 0.989 0.220 0.287 full spectrum 0.993 0.990 0.202 0.275 O O 1100-1250 0.989 0.978 0.262 0.373 1600-1800 0.991 0.987 0.242 0.290 1100-1250, 1600-1800 0.992 0.990 0.220 0.270 full spectrum 0.993 0.990 0.202 0.253 O C-S C 1100-1250 0.988 0.987 0.282 0.304 1600-1800 0.994 0.993 0.184 0.205 O 1100-1250, 1600-1800 0.993 0.991 0.212 0.234 full spectrum 0.989 0.989 0.264 0.285 S 1100-1250 0.988 0.987 0.282 0.305 1600-1800 0.994 0.994 0.184 0.205 1100-1250, 1600-1800 0.993 0.993 0.212 0.230 full spectrum 0.999 0.998 0.081 0.105 O O-S O 1100-1250 0.989 0.974 0.270 0.460 1600-1800 0.993 0.992 0.212 0.232 1100-1250, 1600-1800 0.989 0.989 0.266 0.283 full spectrum 0.998 0.997 0.098 0.127 O S 1100-1250 0.989 0.970 0.270 0.456 1600-1800 0.993 0.992 0.212 0.229 1100-1250, 1600-1800 0.989 0.988 0.266 0.287 full spectrum 0.998 0.997 0.098 0.126 O

This is because the type and composition ratio of fatty acids composing waste cooking oil are different, so the composition ratio of the C-H harmonic band and the C-H bond band according to the molecular structure is also different. It is judged that the prediction performance of the model was the best because the entire wavelength was used.

The average values of the developed models were R ² _cal = 0.987, R ² _val = 0.979, and RMSEC = 0.250, which was confirmed to have high linearity and low prediction error. Among the generated PLSR models, the predictive model with the best performance is a full spectrum model for predicting the amount of S in mixed CS, showing R ² _val=0.998 and RMSEV=0.105.

As an example of the verification result in FIG. 4 , the PLSR model predicts the amount of waste olive oil in the waste oil mixture (O-S) in a static state (A) and the PLSR model predicts the amount of waste olive oil in the waste oil mixture (O-S) in the static state (A) The verification result (B) is shown.

(7) Analysis result - 4

The generated PLSR models consist of 12 models that predict the amount of a specific raw material in a sample mixed with three types of waste cooking oil. The performance of the model was compared using R ² , which is the correlation coefficient, and RMSE (Root Mean Square Error), which is the difference between the actual value and the value predicted by the model.

The results of the PLSR model predicting the amount of each waste cooking oil (WCO) component are shown in Tables 3 and 4 below. As shown in Tables 3 and 4 below, it can be seen that the wavelength range of the PLSR model showing the highest performance according to the type of mixture is varied.

Among the generated PLSR models, the best performing predictive model is the 1600-1800 nm region model for predicting the amount of O in the mixed BOS, and R ² _val=0.994, RMSEV=0.144.

WCO
Mixture WCO Component Wavelength [nm] R ² _cal R ² _val RMSEC RMSEV decision
Model B-C-S B 1100-1250 0.976 0.967 0.304 0.365 1600-1800 0.987 0.986 0.216 0.233 1100-1250, 1600-1800 0.989 0.988 0.205 0.216 full spectrum 0.993 0.992 0.161 0.174 O C 1100-1250 0.950 0.949 0.437 0.448 1600-1800 0.985 0.985 0.234 0.243 O 1100-1250, 1600-1800 0.984 0.983 0.242 0.254 full spectrum 0.985 0.978 0.238 0.292 S 1100-1250 0.973 0.972 0.321 0.333 1600-1800 0.989 0.988 0.199 0.213 O 1100-1250, 1600-1800 0.984 0.983 0.243 0.257 full spectrum 0.983 0.980 0.256 0.269 B-C-O B 1100-1250 0.910 0.862 0.589 0.734 1600-1800 0.959 0.930 0.395 0.519 1100-1250, 1600-1800 0.987 0.978 0.224 0.293 full spectrum 0.994 0.990 0.147 0.195 O C 1100-1250 0.725 0.556 1.032 1.324 1600-1800 0.828 0.709 0.816 1.072 1100-1250, 1600-1800 0.968 0.918 0.352 0.566 full spectrum 0.988 0.974 0.215 0.317 O O 1100-1250 0.934 0.888 0.504 0.661 1600-1800 0.950 0.919 0.440 0.567 1100-1250, 1600-1800 0.982 0.969 0.263 0.346 full spectrum 0.991 0.984 0.186 0.249 O

WCO
Mixture WCO Component Wavelength [nm] R ² _cal R ² _val RMSEC RMSEV decision
Model B-O-S B 1100-1250 0.977 0.976 0.293 0.301 1600-1800 0.985 0.984 0.236 0.252 1100-1250, 1600-1800 0.991 0.990 0.184 0.198 O full spectrum 0.988 0.986 0.214 0.231 O 1100-1250 0.976 0.962 0.300 0.386 1600-1800 0.994 0.994 0.140 0.144 O 1100-1250, 1600-1800 0.994 0.993 0.149 0.156 full spectrum 0.990 0.990 0.189 0.196 S 1100-1250 0.967 0.947 0.357 0.453 1600-1800 0.982 0.980 0.259 0.276 1100-1250, 1600-1800 0.988 0.987 0.214 0.227 O full spectrum 0.987 0.985 0.220 0.239 C-O-S C 1100-1250 0.956 0.952 0.411 0.432 1600-1800 0.976 0.959 0.302 0.397 1100-1250, 1600-1800 0.991 0.985 0.180 0.242 O full spectrum 0.985 0.982 0.241 0.266 O 1100-1250 0.952 0.910 0.428 0.587 1600-1800 0.981 0.978 0.265 0.295 1100-1250, 1600-1800 0.991 0.990 0.178 0.192 full spectrum 0.993 0.992 0.153 0.169 O S 1100-1250 0.943 0.900 0.469 0.632 1600-1800 0.985 0.981 0.239 0.265 1100-1250, 1600-1800 0.988 0.986 0.210 0.230 O full spectrum 0.985 0.983 0.237 0.262

(8) Analysis result - 5

The generated PLSR models consist of 12 models that predict the amount of a specific raw material in a sample mixed with 4 types of waste cooking oil. The performance of the model was compared using R ² , which is the correlation coefficient, and RMSE (Root Mean Square Error), which is the difference between the actual value and the value predicted by the model.

The results of the PLSR model predicting the amount of each waste cooking oil (WCO) component are shown in Table 5 below. As shown in Table 5 below, it can be seen that the wavelength range of the PLSR model showing the highest performance according to the type of mixture is varied.

Among the generated PLSR models, the best performing prediction model is a full spectrum domain model for predicting the amount of O in the mixed BCOS, and it showed R ² _val=0.963 and RMSEV=0.286.

WCO
Mixture WCO Component Wavelength [nm] R ² _cal R ² _val RMSEC RMSEV decision
Model B-C-O-S B 1100-1250 0.761 0.687 0.732 0.841 1600-1800 0.761 0.597 0.732 0.955 1100-1250, 1600-1800 0.887 0.772 0.503 0.717 full spectrum 0.968 0.922 0.264 0.418 O C 1100-1250 0.802 0.762 0.666 0.733 1600-1800 0.835 0.793 0.608 0.683 1100-1250, 1600-1800 0.846 0.814 0.856 0.646 full spectrum 0.946 0.891 0.348 0.496 O O 1100-1250 0.938 0.927 0.372 0.405 1600-1800 0.961 0.949 0.293 0.337 1100-1250, 1600-1800 0.971 0.961 0.251 0.295 full spectrum 0.970 0.963 0.257 0.286 O S 1100-1250 0.749 0.671 0.742 0.857 1600-1800 0.690 0.594 0.825 0.949 1100-1250, 1600-1800 0.886 0.757 0.500 0.732 full spectrum 0.969 0.916 0.279 0.414 O

The developed PLSR model showed the performance of the model with high linearity and low prediction error. The application of near-infrared spectroscopy and PLSR model to the biodiesel production process for biodiesel production is expected to greatly contribute to the improvement of biodiesel yield.

Features, structures, effects, etc. exemplified in each of the above-described embodiments may be combined or modified for other embodiments by those of ordinary skill in the art to which the embodiments belong. Accordingly, the contents related to such combinations and modifications should be interpreted as being included in the scope of the present invention.

Claims (6)

A data collection step (S1) of mixing at least two or more types of waste cooking oil to collect near-infrared absorption spectrum data for each concentration of the mixed raw material;
a modeling step (S2) of obtaining a prediction candidate model according to a wavelength region using a multivariate statistical analysis (Partial Least Squares Analysis; PLSA) method from the obtained near-infrared absorption spectrum data;
With respect to the obtained prediction candidate model, the prediction candidate model of the wavelength region having the highest prediction performance is predicted using any one or more of a coefficient of determination (R ² ) and a root mean square error (RMSE). a model determination step (S3) of determining a model; and
After measuring the near-infrared absorption spectrum of the mixed waste cooking oil sample to be analyzed, the analysis step (S4) of predicting the content of a specific raw material by substituting it into the prediction model of the determined wavelength region; .
According to claim 1,
In the modeling step (S2), the wavelength region is a first region from 898.77 nm to 2132.65 nm, a second region from 1,100 nm to 1250 nm, a third region from 1,600 nm to 1800 nm, and a third region from 1,100 nm to 1250 nm and 1,600 nm to 1800 nm. A raw material mixing ratio analysis method of waste cooking oil, characterized in that it includes 4 areas.
According to claim 1,
The model determining step (S3) is a step of determining a prediction candidate model in the wavelength region having the highest prediction performance compared with the actual value among the prediction candidate models according to the wavelength region as the prediction model. Mixing ratio analysis method.
According to claim 1,
The model determining step (S3) is a prediction candidate model of the first region from 898.77 nm to 2132.65 nm in order to predict the content of the waste sunflower oil when the waste cooking oil is a mixture of sunflower oil and canola oil. A method for analyzing the raw material mixing ratio of waste cooking oil, characterized in that it is the step of determining as a predictive model.
According to claim 1,
The model determination step (S3) is a third of 1600 nm to 1800 nm in order to predict the content of the waste olive oil when the waste cooking oil is a mixture of waste soybean oil, waste olive oil, and waste sunflower oil Raw material mixing ratio analysis method of waste cooking oil, characterized in that the step of determining the prediction candidate model of the region as the prediction model.
According to claim 1,
The model determining step (S3) is to predict the content of waste olive oil when the waste cooking oil is a mixture of waste soybean oil, waste olive oil, waste canola oil and sunflower oil. A method for analyzing the raw material mixing ratio of waste cooking oil, characterized in that the step of determining the prediction candidate model of the first region of 898.77 nm to 2132.65 nm as the prediction model for the purpose.

Images