Biomass pyrolysis product prediction method based on machine learning model

Abstract

The invention discloses a biomass pyrolysis product prediction method based on a machine learning model, and the method comprises the steps: collecting data under different biomass pyrolysis conditions, and carrying out the processing of the data, and obtaining target data; constructing a machine learning model, and training the machine learning model based on the target data to obtain a target machine learning model; according to the target machine learning model, predicting the yield of the biochar and the content of nitrogen, phosphorus and potassium, and obtaining a prediction result of the yield of the biochar and the composition of nitrogen, phosphorus and potassium elements. By introducing the machine learning model and combining the data enhancement technology, high-precision prediction can be performed on the yield of the biochar and the contents of nitrogen, phosphorus and potassium of the biochar based on the element composition and pyrolysis process parameters of the biomass raw material, and technical support is provided for optimization design of the biomass pyrolysis process and application of the biochar fertilizer.

Description

Technical Field

[0001] The present invention belongs to the field of biomass pyrolysis and biochar fertilizer, and in particular relates to a method for predicting biomass pyrolysis products based on a machine learning model. Background Art

[0002] In recent years, biomass pyrolysis technology has been widely used in renewable energy and agricultural waste treatment. Biochar, as one of the pyrolysis products, is considered to be a material with multiple ecological benefits, especially showing great potential in improving soil quality and carbon fixation. At the same time, nitrogen, phosphorus and potassium (NPK) in biochar are the three main nutrients required for plant growth, making it widely used as a carbon-based fertilizer in agriculture.

[0003] However, the complexity of biomass raw materials and the diversity of pyrolysis processes make it difficult to accurately predict biochar yield and its NPK content. Traditional prediction methods mostly rely on experimental measurements and empirical formulas, which are inefficient and the results are not universal. In addition, there are many factors that affect the characteristics of biochar, including biomass type, pyrolysis temperature, reaction time, etc. Therefore, a technical means that can process multidimensional complex data is needed to accurately predict biochar yield and its NPK content.

[0004] At present, the research on the prediction of biomass pyrolysis products using machine learning methods is relatively limited, and the research on the migration and distribution characteristics of NPK is not in-depth enough, lacking a systematic and accurate prediction model. Summary of the invention

[0005] The purpose of the present invention is to provide a method for predicting the yield and NPK content of biochar from biomass pyrolysis based on machine learning, so as to solve the problems of prediction accuracy and efficiency of traditional methods. By introducing a machine learning model and combining data enhancement technology, the yield of biochar and its nitrogen, phosphorus and potassium contents can be predicted with high accuracy based on the elemental composition of biomass raw materials and pyrolysis process parameters, providing technical support for the optimal design of biomass pyrolysis process and the application of biochar fertilizer.

[0006] To achieve the above object, the present invention provides a method for predicting biomass pyrolysis products based on a machine learning model, comprising:

[0007] Collecting data under different biomass pyrolysis conditions, and processing the data to obtain target data;

[0008] Constructing a machine learning model, and training the machine learning model based on the target data to obtain a target machine learning model;

[0009] The biochar production and its nitrogen, phosphorus and potassium contents are predicted according to the target machine learning model to obtain the prediction results of the biochar yield and its nitrogen, phosphorus and potassium element composition.

[0010] Preferably, the data under different biomass pyrolysis conditions include different biomass types and their elemental compositions and pyrolysis conditions.

[0011] Preferably, the biomass types include but are not limited to straw, wood chips, and livestock and poultry manure;

[0012] The element composition includes but is not limited to hydrogen, oxygen, nitrogen, phosphorus, and potassium;

[0013] The pyrolysis conditions include but are not limited to temperature, heating rate, pyrolysis time and feeding rate.

[0014] Preferably, processing the data includes but is not limited to data cleaning, missing value filling and standardization;

[0015] To address the problem of insufficient data, data augmentation technology is used to expand the dataset by adding noise or generating synthetic data.

[0016] Preferably, the data cleaning process includes:

[0017] For the detection and processing of outliers, the box plot analysis method was used to eliminate obvious outliers;

[0018] The missing value filling process includes:

[0019] For the processing of missing data, the missing values ​​are filled by filling the mean value or using the interpolation algorithm;

[0020] The standardized process includes:

[0021] For data standardization, data of different dimensions are standardized to a unified range.

[0022] Preferably, the process of training the machine learning model based on the target data includes:

[0023] The yield and elemental composition of biochar are predicted based on multidimensional input variables of biomass elemental composition and pyrolysis parameters. Cross-validation technology is used, and the model performance is evaluated by evaluation indicators. The model hyperparameters are tuned according to the model performance to obtain the target machine learning model.

[0024] Preferably, the cross-validation technique is a 5-fold cross-validation method;

[0025] The evaluation indicators include but are not limited to mean square error, root mean square error and determination coefficient.

[0026] Preferably, the machine learning model includes but is not limited to XGBoost, random forest, support vector regression, and BP-ANN.

[0027] Preferably, the method further comprises:

[0028] The target machine learning model was interpreted by the Shapley value analysis method to identify the key factors affecting biochar yield and NPK content, and further optimize the pyrolysis process design.

[0029] Preferably, the formula expression of the target machine learning model is:

[0030]

[0031] in, To predict biochar yield or element content, x i is the input feature, w i is the weight of each feature in the XGBoost model, and b is the bias term.

[0032] Compared with the prior art, the present invention has the following advantages and technical effects:

[0033] The method for predicting biomass pyrolysis products based on a machine learning model provided by the present invention solves the problem that traditional methods cannot accurately predict biochar yield and NPK content by introducing multiple machine learning algorithms and data enhancement techniques. This method can not only perform generalized predictions under multiple biomass types and pyrolysis conditions, but also quantitatively analyze the migration dynamics of key elements such as N, P, and K, providing a new idea for optimizing the pyrolysis process and improving the fertilizer efficiency of biochar. BRIEF DESCRIPTION OF THE DRAWINGS

[0034] The drawings constituting a part of the present application are used to provide a further understanding of the present application. The illustrative embodiments and descriptions of the present application are used to explain the present application and do not constitute an improper limitation on the present application. In the drawings:

[0035] Figure 1 The figure is a schematic diagram of a method flow of an embodiment of the present invention. DETAILED DESCRIPTION

[0036] It should be noted that, in the absence of conflict, the embodiments and features in the embodiments of the present application can be combined with each other. The present application will be described in detail below with reference to the accompanying drawings and in combination with the embodiments.

[0037] It should be noted that the steps shown in the flowcharts of the accompanying drawings can be executed in a computer system such as a set of computer executable instructions, and that, although a logical order is shown in the flowcharts, in some cases, the steps shown or described can be executed in an order different from that shown here.

[0038] like Figure 1 As shown, this embodiment provides a method for predicting biomass pyrolysis products based on a machine learning model, comprising:

[0039] Collect data under different biomass pyrolysis conditions and process the data to obtain target data;

[0040] Build a machine learning model, train the machine learning model based on target data, and obtain a target machine learning model;

[0041] The biochar production and its nitrogen, phosphorus, and potassium contents were predicted according to the target machine learning model, and the prediction results of biochar yield and its nitrogen, phosphorus, and potassium element composition were obtained.

[0042] Furthermore, data collection: collect elemental composition data of different biomass types, including carbon (C), hydrogen (H), oxygen (O), nitrogen (N), phosphorus (P), potassium (K) content and pyrolysis process parameters, such as pyrolysis temperature, heating rate, pyrolysis time and feeding rate.

[0043] More specifically, the data collected in this embodiment covers experimental data under various biomass pyrolysis conditions, including different biomass types, their elemental compositions, and pyrolysis conditions. Among them, biomass types include straw, sawdust, livestock and poultry manure, etc.; elemental compositions include hydrogen, oxygen, nitrogen, phosphorus, potassium, etc.; pyrolysis conditions include temperature, heating rate, pyrolysis time, and feeding rate, etc. By collecting the above data, the input features for training the machine learning model are obtained.

[0044] Further, data preprocessing: To ensure the training effect of the model, the following preprocessing steps are required for the data:

[0045] The collected data is processed, including data cleaning, missing value filling and standardization. For the problem of insufficient data, data enhancement technology is used to expand the data set by adding noise or generating synthetic data to improve the generalization ability of the model.

[0046] Specifically, missing data is processed by filling the mean value or using an interpolation algorithm to complete the missing values. For the detection and processing of outliers, the box plot analysis method is used to remove obvious outliers. For data standardization, data of different dimensions are standardized to a unified range to improve the stability of model training.

[0047] Further, model construction: a prediction model is constructed using a machine learning algorithm to predict the yield of biochar and its nitrogen (N), phosphorus (P), and potassium (K) content based on multidimensional input variables. The machine learning model used in this embodiment may include random forest, XGBoost, support vector regression (SVR), BP-ANN, etc., focusing on predicting the yield of biochar and the migration of nitrogen, phosphorus, and potassium elements.

[0048] Furthermore, model training and optimization: the model was trained using cross-validation techniques, and mean square error (MSE), root mean square error (RMSE) and coefficient of determination (R 2 ) and other indicators to evaluate model performance. Hyperparameters of the model are tuned to improve prediction accuracy.

[0049] Further, model interpretation and feature importance analysis: The model was interpreted through Shapley value analysis (SHAP), the key factors affecting biochar yield and NPK content were identified, and the pyrolysis process design was further optimized. The specific steps are as follows:

[0050] Model selection: Based on the amount of data, feature dimension and prediction target, the XGBoost model is selected as the main algorithm.

[0051] Model training: The biomass elemental composition and pyrolysis conditions are used as input features to predict the yield and elemental composition of biochar. A 5-fold cross validation is used during model training, and the mean square error (MSE), root mean square error (RMSE) and R 2 The value is used as the model evaluation indicator.

[0052] Data augmentation: To further improve the generalization ability of the model, data augmentation techniques are applied, such as expanding the training set by adding noise or generating synthetic data.

[0053] Through XGBoost model training, the prediction results of biochar yield and its nitrogen, phosphorus, and potassium composition were obtained. 2 The value is above 0.95, indicating that it has a good fitting effect. In addition, the present invention also explains the influence of input characteristics (such as temperature, heating rate, etc.) on the prediction results through the SHAP (SHapley Additive exPlanations) analysis method, providing data support for the optimization of biomass pyrolysis process.

[0054] Furthermore, predictive application: the optimized model is used in the actual biomass pyrolysis process. Based on the input biomass composition and pyrolysis parameters, the biochar yield and its NPK content are accurately predicted to provide data support for practical applications.

[0055] In specific applications, users can input the basic composition of biomass and pyrolysis conditions through a graphical user interface (GUI). The machine learning model of this embodiment can predict the yield of biochar and its elemental composition, especially the migration of nitrogen, phosphorus and potassium elements, in real time. This provides an important reference for the application of biochar as fertilizer. The following are the main steps to implement this method:

[0056] Step 1: The user inputs the elemental composition of the biomass (C, H, O, N, P, K) and parameters such as pyrolysis temperature and heating rate.

[0057] Step 2: The system calls the trained XGBoost model to predict biochar yield and elemental composition.

[0058] Step 3: Output the prediction results and display the characteristics and prediction errors of biochar through a visual interface.

[0059] The present invention realizes accurate prediction of biochar characteristics by mining and training models on a large amount of public literature data, and provides a basis for evaluating agricultural biochar, which has important application value.

[0060] More specifically, the prediction model is expressed as:

[0061]

[0062] in, To predict biochar yield or element content, x i is the input feature, w i is the weight of each feature in the XGBoost model, and b is the bias term.

[0063] This embodiment can also use other machine learning models (such as neural networks, decision trees, etc.) instead of XGBoost models for prediction, and select appropriate data enhancement methods according to different biomass sources. In addition, the migration efficiency of NPK can be directly improved by optimizing pyrolysis process parameters.

[0064] This example uses machine learning technology to predict biochar yield and its nitrogen (N), phosphorus (P), and potassium (K) content, improves the generalization ability of the model through data enhancement, uses feature importance analysis methods to reveal the key factors affecting NPK migration and distribution, and optimizes the pyrolysis process design, solving the problem that traditional methods cannot accurately predict biochar yield and NPK content. This method can not only perform generalized predictions under a variety of biomass types and pyrolysis conditions, but also quantitatively analyze the migration dynamics of key elements such as N, P, and K, providing new ideas for optimizing pyrolysis processes and improving biochar fertilizer efficiency.

[0065] The above are only preferred specific implementations of the present application, but the protection scope of the present application is not limited thereto. Any changes or substitutions that can be easily thought of by a person skilled in the art within the technical scope disclosed in the present application should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be based on the protection scope of the claims.

Claims

1. A method for predicting biomass pyrolysis products based on a machine learning model, characterized in that: include: Collecting data under different biomass pyrolysis conditions, and processing the data to obtain target data; Constructing a machine learning model, and training the machine learning model based on the target data to obtain a target machine learning model; The biochar production and its nitrogen, phosphorus and potassium contents are predicted according to the target machine learning model to obtain the prediction results of the biochar yield and its nitrogen, phosphorus and potassium element composition.

2. The method according to claim 1, characterized in that The data under different biomass pyrolysis conditions include different biomass types, elemental compositions and pyrolysis conditions.

3. The method according to claim 2, characterized in that The types of biomass include but are not limited to straw, wood chips, and livestock and poultry manure; The element composition includes but is not limited to hydrogen, oxygen, nitrogen, phosphorus, and potassium; The pyrolysis conditions include but are not limited to temperature, heating rate, pyrolysis time and feeding rate.

4. The method according to claim 1, characterized in that: Processing the data includes but is not limited to data cleaning, missing value filling and standardization; To address the problem of insufficient data, data augmentation technology is used to expand the dataset by adding noise or generating synthetic data.

5. The method according to claim 4, characterized in that The data cleaning process includes: For the detection and processing of outliers, the box plot analysis method was used to eliminate obvious outliers; The missing value filling process includes: For the processing of missing data, the missing values ​​are filled by filling the mean value or using the interpolation algorithm; The standardized process includes: For data standardization, data of different dimensions are standardized to a unified range.

6. The method according to claim 1, characterized in that The process of training the machine learning model based on the target data includes: The yield and elemental composition of biochar are predicted based on multidimensional input variables of biomass elemental composition and pyrolysis parameters. Cross-validation technology is used, and the model performance is evaluated by evaluation indicators. The model hyperparameters are tuned according to the model performance to obtain the target machine learning model.

7. The method according to claim 6, characterized in that The cross-validation technique is a 5-fold cross-validation method; The evaluation indicators include but are not limited to mean square error, root mean square error and determination coefficient.

8. The method according to claim 1, characterized in that: The machine learning models include but are not limited to XGBoost, random forest, support vector regression, and BP-ANN.

9. The method according to claim 1, characterized in that: The method further comprises: The target machine learning model was interpreted by the Shapley value analysis method to identify the key factors affecting biochar yield and NPK content, and further optimize the pyrolysis process design.

10. The method according to claim 1, characterized in that The formula expression of the target machine learning model is: in, To predict biochar yield or element content, x i is the input feature, w i is the weight of each feature in the XGBoost model, and b is the bias term.

Images