Biodiesel quality prediction method based on machine learning

Abstract

According to the biodiesel quality prediction method based on machine learning, a data acquisition module acquires grease samples, the data acquisition module transmits the acquired grease samples to a preprocessing module, the preprocessing module preprocesses original data, and the data preprocessing module preprocesses the grease samples and then sends the preprocessed grease samples to a feature extraction module. And the feature analysis module screens out features related to the grease quality from the preprocessed data through correlation analysis to obtain a grease feature vector xi. The data processing module divides the quality of the historical grease data into three quality grades of A, B and C according to expert experience; a data processing module selects m groups of grease feature vectors xi from a database, the data processing module forms a training data set by the m groups of grease feature vectors xi and corresponding quality labels, and the data processing module respectively constructs a first classifier fAB (x), a second classifier fAC (x) and a third classifier fBC (x). According to the method, the prediction model is established, and the prediction model can rapidly predict the grease quality of a new sample in real time.

Description

Technical Field

[0001] The present invention relates to the field of biodiesel, and in particular to a biodiesel quality prediction method based on machine learning. Background Art

[0002] Biodiesel is a clean fuel produced by chemical conversion of organic resources such as vegetable oil or animal fat. The raw materials of biodiesel are converted from waste such as vegetable oil and waste cooking oil, and have strong biodegradability. The quality of biodiesel is a key indicator. The quality of biodiesel not only affects its combustion efficiency and tail gas emissions, but also affects its storage stability and equipment compatibility. Therefore, it is necessary to monitor the quality of biodiesel regularly. At present, the quality detection of biodiesel is mainly carried out by offline physical and chemical index detection of key parameters such as acid value, iodine value, and fatty acid composition. This detection method has the following shortcomings:

[0003] Offline physical and chemical index detection relies on a single indicator to judge the quality of biodiesel. Although indicators such as acid value and iodine value can reflect the quality of biodiesel in some cases, these indicators cannot reflect the quality of oil products well. At the same time, offline detection has a long cycle and slow response, and it is impossible to achieve real-time detection and feedback of the production process. Since the above detection methods mostly rely on regular sampling and testing, it is difficult to adjust production parameters in time to optimize the quality of biodiesel. Therefore, in order to solve the above problems, a biodiesel quality prediction method based on machine learning is needed. Summary of the invention

[0004] In view of the shortcomings of the prior art, the present invention proposes a biodiesel quality prediction method based on machine learning. The specific technical scheme is as follows:

[0005] A biodiesel quality prediction method based on machine learning, characterized in that:

[0006] The steps include:

[0007] S1: the data acquisition module collects grease samples;

[0008] S2: the data acquisition module transmits the collected grease sample to the preprocessing module;

[0009] S3: the preprocessing module preprocesses the raw data, and the data preprocessing module sends the preprocessed grease sample to the feature extraction module;

[0010] S4: The feature analysis module selects features related to oil quality from the preprocessed data through correlation analysis to obtain the oil feature vector x i The oil samples include acid value, density, viscosity, fatty acid composition and production temperature;

[0011] Among them, x i =[x i1 ,x i2 ,...,x in ];

[0012] x in is the feature value, and n is the number of features screened by the feature analysis module;

[0013] S5: The data processing module divides the quality of historical oil and fat data into three quality levels: A, B, and C according to expert experience;

[0014] S6: The data processing module selects m groups of oil feature vectors x in the database. i The data processing module converts m groups of fat feature vectors x i The corresponding quality labels constitute the training data set;

[0015] D={([x i1 ,x i2 ,...,x in ],y i )|y i ∈{A,B,C},i=1,2,...,m}

[0016] Where m is the number of samples;

[0017] S7: The data processing modules construct the first classifier f AB (x), the second classifier f AC (x), the third classifier f BC (x);

[0018] The first classifier f AB (x), the second classifier f AC (x) and the third classifier f BC (x) are all based on the oil characteristic vector x i As input vector;

[0019] The first classifier f AB (x) Output label is quality A / quality B;

[0020] The second classifier f AC (x) Output label is quality A / quality C;

[0021] The third classifier f BC (x) The output label is quality B / quality C;

[0022] S8: The data acquisition module collects new oil samples and processes them through the preprocessing module, and then sends them to the feature extraction module. The feature extraction module extracts the oil feature vector x new ;

[0023] S9: The feature extraction module extracts the grease feature vector x new are respectively input into the first classifier f AB (x), the second classifier f AC (x) and the third classifier f BC (x);

[0024] For the grease feature vector x new :

[0025]

[0026] S10: According to the first classifier f AB (x new ), the second classifier f AC (x new ) and the third classifier f BC (x new ), the data processing module uses a voting method to determine the final category of the new grease sample:

[0027]

[0028] Among them, f i (x) is the output result of the above three classifiers;

[0029] I is the indicator function, if f i (x) The predicted category is k, then I(f i (x)=k)=1, otherwise it is 0.

[0030] In order to better realize the present invention, it is possible to further:

[0031] The data processing module periodically generates a quality report, extracts the oil quality status within a periodic time period, and sends it to the visualization module.

[0032] Further: a data management module is provided, the quality prediction data and the corresponding feature data are archived in the data management module, and the storage format includes sample ID, feature vector, prediction result and timestamp;

[0033] The sample ID corresponds to the unique identifier of the sample.

[0034] The feature vector includes relevant features of each oil sample;

[0035] The predicted quality includes oil quality grade;

[0036] The timestamp includes the predicted time for subsequent trend analysis.

[0037] The beneficial effects of the present invention are as follows: by establishing a prediction model, the prediction model can quickly and in real time predict the oil quality of new samples, which greatly reduces the detection cost compared to physical and chemical determination under laboratory conditions. By setting a feature extraction module, the feature extraction module can extract multiple feature vectors and mine the hidden relationship between features through a support vector machine, thereby improving the prediction accuracy. By setting up an abnormal detection module and cooperating with the quality prediction model, it is possible to synchronously detect abnormal behaviors of samples in the feature space. It is ensured that even if a specific quality label is predicted, potential abnormal samples can be identified, thereby improving the reliability and accuracy of the system. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] Figure 1 It is a work flow chart of the present invention;

[0039] Figure 2 It is a system diagram of the present invention. DETAILED DESCRIPTION

[0040] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0041] like Figure 1 and Figure 2 As shown:

[0042] A biodiesel quality prediction method based on machine learning comprises the following steps:

[0043] S1: the data acquisition module collects grease samples;

[0044] S2: the data acquisition module transmits the collected grease sample to the preprocessing module;

[0045] S3: the preprocessing module preprocesses the raw data, and the data preprocessing module sends the preprocessed grease sample to the feature extraction module;

[0046] S4: The feature analysis module selects features related to oil quality from the preprocessed data through correlation analysis to obtain the oil feature vector x i The oil samples include acid value, density, viscosity, fatty acid composition and production temperature;

[0047] Among them, x i =[x i1 ,x i2 ,...,x in ];

[0048] x in is the feature value, and n is the number of features screened by the feature analysis module;

[0049] S5: The data processing module divides the quality of historical oil and fat data into three quality levels: A, B, and C according to expert experience;

[0050] S6: The data processing module selects m groups of oil feature vectors x in the database. i The data processing module converts m groups of fat feature vectors x i The corresponding quality labels constitute the training data set;

[0051] D={([x i1 ,x i2 ,...,x in ],y i )|y i ∈{A,B,C},i=1,2,...,m}

[0052] Where m is the number of samples;

[0053] S7: The data processing modules construct the first classifier f AB (x), the second classifier f AC (x), the third classifier f BC (x), the first classifier f AB (x), the second classifier f AC (x) and the third classifier f BC (x) All use support vector machines;

[0054] The first classifier f AB (x), the second classifier f AC (x) and the third classifier f BC (x) are all based on the oil characteristic vector x i As input vector;

[0055] The first classifier f AB (x) Output label is quality A / quality B;

[0056] The second classifier f AC (x) Output label is quality A / quality C;

[0057] The third classifier f BC (x) The output label is quality B / quality C;

[0058] S8: The data acquisition module collects new oil samples and processes them in the preprocessing module, and then sends them to the feature extraction module. The feature extraction module extracts the oil feature vector x new ;

[0059] S9: An abnormality detection module is provided, and the feature extraction module extracts the grease feature vector x new are respectively input into the first classifier f AB (x), the second classifier f AC (x) and the third classifier f BC (x).

[0060] The fat feature vector x new At the same time, it is input into the anomaly detection module, which is used to analyze the deviation of the grease feature vector in the quality feature space and issue an alarm for abnormal data.

[0061] For the grease feature vector x new :

[0062]

[0063] S10: According to the first classifier f AB (x new ), the second classifier f AC (x new ) and the third classifier f BC (x new ), the data processing module uses a voting method to determine the final category of the new grease sample:

[0064]

[0065] Among them, f i (x) is the output result of the above three classifiers;

[0066] I is the indicator function, if f i (x) The predicted category is k, then I(f i (x)=k)=1, otherwise it is 0.

[0067] S11: A visualization module and a data management module are provided. The data processing module periodically generates a quality report, extracts the oil quality situation within the periodic time period, and sends it to the visualization module. The quality prediction data and the corresponding feature data are archived in the data management module. The storage format includes sample ID, feature vector, prediction result and timestamp. The sample ID corresponds to the unique identifier of the sample, and the feature vector includes the relevant features of each oil sample. The predicted quality includes the oil quality grade, and the timestamp includes the predicted time, which is used for subsequent trend analysis.

[0068] It will be apparent to those skilled in the art that the invention is not limited to the details of the exemplary embodiments described above and that the invention can be implemented in other specific forms without departing from the spirit or essential features of the invention. Therefore, the embodiments should be considered exemplary and non-limiting in all respects, and the scope of the invention is defined by the appended claims rather than the foregoing description, and it is intended that all variations falling within the meaning and scope of the equivalent elements of the claims be included in the invention. Any reference numeral in a claim should not be considered as limiting the claim to which it relates.

[0069] In addition, it should be understood that although the present specification is described according to implementation modes, not every implementation mode contains only one independent technical solution. This description of the specification is only for the sake of clarity. Those skilled in the art should regard the specification as a whole. The technical solutions in each embodiment may also be appropriately combined to form other implementation modes that can be understood by those skilled in the art.

Claims

1. A biodiesel quality prediction method based on machine learning, characterized in that: The steps include: S1: the data acquisition module collects grease samples; S2: the data acquisition module transmits the collected grease sample to the preprocessing module; S3: the preprocessing module preprocesses the raw data, and the data preprocessing module sends the preprocessed grease sample to the feature extraction module; S4: The feature analysis module selects features related to oil quality from the preprocessed data through correlation analysis to obtain the oil feature vector x i The oil samples include acid value, density, viscosity, fatty acid composition and production temperature; Among them, x i =[x i1 ,x i2 ,...,x in ]; x in is the feature value, and n is the number of features screened by the feature analysis module; S5: The data processing module divides the quality of historical oil and fat data into three quality levels: A, B, and C according to expert experience; S6: The data processing module selects m groups of oil feature vectors x in the database. i The data processing module converts m groups of fat feature vectors x i The corresponding quality labels constitute the training data set; D={([x i1 ,x i2 ,...,x in ],y i )|y i ∈{A,B,C},i=1,2,...,m} Where m is the number of samples; S7: The data processing modules construct the first classifier f AB (x), the second classifier f AC (x), the third classifier f BC (x); The first classifier f AB (x), the second classifier f AC (x) and the third classifier f BC (x) are all based on the oil characteristic vector x i As input vector; The first classifier f AB (x) Output label is quality A / quality B; The second classifier f AC (x) Output label is quality A / quality C; The third classifier f BC (x) The output label is quality B / quality C; S8: The data acquisition module collects new oil samples and processes them through the preprocessing module, and then sends them to the feature extraction module. The feature extraction module extracts the oil feature vector x new ; S9: The feature extraction module extracts the grease feature vector x new are respectively input into the first classifier f AB (x), the second classifier f AC (x) and the third classifier f BC (x); For the grease feature vector x new : S10: According to the first classifier f AB (x new ), the second classifier f AC (x new ) and the third classifier f BC (x new ), the data processing module uses a voting method to determine the final category of the new grease sample: Among them, f i (x) is the output result of the above three classifiers; I is the indicator function, if f i (x) The predicted category is k, then I(f i (x)=k)=1, otherwise it is 0.

2. The method for predicting biodiesel quality based on machine learning according to claim 1, characterized in that: The data processing module periodically generates a quality report, extracts the oil quality status within a periodic time period, and sends it to the visualization module.

3. The method for predicting biodiesel quality based on machine learning according to claim 2, characterized in that: An abnormality detection module is provided, and the grease feature vector x new At the same time, it is input into the anomaly detection module, which is used to analyze the deviation of the grease feature vector in the quality feature space and issue an alarm for abnormal data.

4. The method for predicting biodiesel quality based on machine learning according to claim 3, characterized in that: A data management module is provided, and the quality prediction data and the corresponding feature data are archived in the data management module, and the storage format includes sample ID, feature vector, prediction result and timestamp; The sample ID corresponds to the unique identifier of the sample. The feature vector includes relevant features of each oil sample; The predicted quality includes oil quality grade; The timestamp includes the predicted time for subsequent trend analysis.

Images