Prediction method and device of biomass pyrolysis product, computer equipment and storage medium

Abstract

The invention provides a method and device for predicting biomass pyrolysis products, computer equipment and a storage medium, and belongs to the field of pyrolysis product prediction.The method comprises the steps that historical data of biomass to be predicted are obtained; determining a first prediction model and a second prediction model with minimum prediction error values in the plurality of models to be selected according to the historical data; obtaining a current characteristic parameter and a current pyrolysis parameter of the to-be-predicted biomass; inputting the current characteristic parameters into a first prediction model, and obtaining predicted carbon content, hydrogen content, nitrogen content and oxygen content through the first prediction model; and inputting the predicted carbon content, hydrogen content, nitrogen content and oxygen content and the current pyrolysis parameters into a second prediction model, and obtaining a pyrolysis product of the to-be-predicted biomass through the second model. The method can improve the prediction accuracy of the pyrolysis product.

Description

Technical Field

[0001] The present invention belongs to the field of pyrolysis product prediction, and specifically relates to a method, device, computer equipment and storage medium for predicting biomass pyrolysis products. Background Art

[0002] Biomass pyrolysis is the process of decomposing biomass into gas, liquid and solid products by heating under anaerobic or micro-aerobic conditions. Biomass absorbs heat during the heating process and the temperature gradually rises. When the temperature reaches a certain level, the biomass begins to decompose and generate gas, liquid and solid products.

[0003] In recent years, biomass pyrolysis technology has received widespread attention and made significant research progress. Predicting the products of biomass pyrolysis can further optimize the biomass pyrolysis process, thereby improving the yield and quality of the products.

[0004] At present, machine learning algorithms are mainly used to build a single prediction model to predict the composition of biomass pyrolysis products. However, the models built by different machine learning algorithms have different prediction results for biomass pyrolysis products under different circumstances. Therefore, the above method has low accuracy in predicting pyrolysis products.

[0005] Therefore, how to improve the prediction accuracy of biomass pyrolysis products becomes a problem to be solved. Summary of the invention

[0006] In order to solve the problem of low prediction accuracy of biomass pyrolysis products in the prior art, the present invention provides a method for predicting biomass pyrolysis products.

[0007] In order to achieve the above object, the present invention provides the following technical solutions:

[0008] A method for predicting biomass pyrolysis products, characterized by comprising:

[0009] Acquire historical data of the biomass to be predicted; wherein the historical data includes historical characteristic parameters of the biomass to be predicted, historical pyrolysis parameters, historical pyrolysis products, and carbon content, hydrogen content, nitrogen content, and oxygen content corresponding to the historical characteristic parameters;

[0010] Determine a first prediction model and a second prediction model with the smallest prediction error value from a plurality of models to be selected according to the historical data; wherein the first prediction model is obtained by training with the historical characteristic parameters and the corresponding carbon content, hydrogen content, nitrogen content and oxygen content as samples; and the second prediction model is obtained by training with the carbon content, hydrogen content, nitrogen content and oxygen content corresponding to the historical characteristic parameters, as well as historical pyrolysis parameters and historical pyrolysis products as samples;

[0011] Acquiring current characteristic parameters and current pyrolysis parameters of the biomass to be predicted;

[0012] Inputting the current characteristic parameters into the first prediction model, and obtaining predicted carbon content, hydrogen content, nitrogen content and oxygen content through the first prediction model;

[0013] The predicted carbon content, hydrogen content, nitrogen content and oxygen content and current pyrolysis parameters are input into the second prediction model, and the pyrolysis product of the biomass to be predicted is obtained through the second model.

[0014] Preferably, the multiple models to be selected include multiple models for predicting carbon content, hydrogen content, nitrogen content and oxygen content, and multiple models for predicting pyrolysis products, and determining the first prediction model and the second prediction model with the smallest prediction error value among the multiple models to be selected according to the historical data includes:

[0015] Inputting the historical characteristic parameters into the multiple models for predicting carbon content, hydrogen content, nitrogen content and oxygen content, and inputting the carbon content, hydrogen content, nitrogen content and oxygen content corresponding to the historical characteristic parameters and the historical pyrolysis parameters into the multiple models for predicting pyrolysis products;

[0016] Predicting the carbon content, hydrogen content and oxygen content corresponding to the historical characteristic parameters to obtain a first result; predicting the pyrolysis product of the biomass to be predicted to obtain a second result;

[0017] Compare the first result with the carbon content, hydrogen content, nitrogen content and oxygen content corresponding to the historical characteristic parameters to determine multiple mean square errors (MSEs), and determine the prediction model corresponding to the minimum value among the multiple mean square errors (MSEs) as the first prediction model;

[0018] The second result is compared with historical pyrolysis products to determine multiple mean square errors (MSEs), and the prediction model corresponding to the minimum value of the multiple mean square errors (MSEs) is determined as the second prediction model.

[0019] Preferably, before inputting the historical characteristic parameters into the multiple models for predicting the carbon content, hydrogen content, nitrogen content and oxygen content, and inputting the carbon content, hydrogen content, nitrogen content and oxygen content corresponding to the historical characteristic parameters and the historical pyrolysis parameters into the multiple models for predicting the pyrolysis products, the method further comprises:

[0020] Performing deduplication preprocessing on the historical data to obtain a first data set;

[0021] Dividing the first data set into a training data set and a test data set;

[0022] Training the plurality of models to be selected by using the training data set;

[0023] The carbon content, hydrogen content and oxygen content corresponding to the historical characteristic parameters are predicted to obtain a first result; and the pyrolysis products of the biomass to be predicted are predicted to obtain a second result, including: predicting the carbon content, hydrogen content and oxygen content corresponding to the historical characteristic parameters in the test data set according to the trained model to obtain the first result, and predicting the pyrolysis products of the biomass to be predicted in the test data set to obtain the second result.

[0024] Preferably, the historical characteristic parameters include moisture content M, ash content ASH, volatile matter VM, and fixed carbon FC, and the training set is used to train a plurality of the models to be selected, including:

[0025] The moisture content M, ash content ASH, volatile matter VM, and fixed carbon FC in the training data set are used as characteristic variables, and the carbon content, hydrogen content, nitrogen content, and oxygen content corresponding to the historical characteristic parameters are used as target variables, and a plurality of models for predicting the carbon content, hydrogen content, nitrogen content, and oxygen content are trained;

[0026] The carbon content, hydrogen content, nitrogen content and oxygen content corresponding to the historical characteristic parameters in the training data set, and the historical pyrolysis parameters are used as characteristic variables, and the historical pyrolysis products are used as target variables to train multiple models for predicting pyrolysis products.

[0027] Preferably, the formula for determining the mean square error MSE is:

[0028]

[0029] Where n is the number of samples, Yi is the i-th true value, and Fi is the i-th predicted value.

[0030] Preferably, the pyrolysis parameters include pyrolysis final temperature FT, heating rate HR, sweep rate FR, particle size PS, and the method further comprises:

[0031] Analyzing the historical data to obtain a histogram of the biomass characteristics to be predicted, a histogram of the pyrolysis conditions, a scatter plot of the biomass characteristics, a scatter plot of the pyrolysis conditions, and a correlation heat map;

[0032] Wherein, the histogram of the biomass characteristics is used to characterize the distribution of the moisture content M, ash content ASH, volatile matter VM, fixed carbon FC, carbon content, hydrogen content, nitrogen content and oxygen content in the biomass to be predicted; the histogram of the pyrolysis conditions is used to characterize the distribution of the final pyrolysis temperature FT, heating rate HR, sweep flow rate FR, and particle size PS; the scatter plot of the biomass characteristics is used to characterize the influence relationship between each of the moisture content M, ash content ASH, volatile matter VM, fixed carbon FC, carbon content, hydrogen content, nitrogen content and oxygen content; the scatter plot of the pyrolysis conditions is used to characterize the influence relationship between each value in the final pyrolysis temperature FT, heating rate HR, sweep flow rate FR, and particle size PS; the correlation heat map is used to characterize the influence relationship between the biomass characteristics and the pyrolysis conditions;

[0033] The pyrolysis process of the biomass to be predicted is optimized according to the histogram of the biomass characteristics to be predicted, the histogram of the pyrolysis conditions, the scatter plot of the biomass characteristics, the scatter plot of the pyrolysis conditions and the correlation heat map.

[0034] Preferably, the multiple models to be selected include a random forest model, a decision tree model, an XGBoost model and a GDBT model.

[0035] The present invention also provides a device for predicting biomass pyrolysis products, characterized in that it comprises:

[0036] A first acquisition module is used to acquire historical data of the biomass to be predicted; wherein the historical data includes historical characteristic parameters of the biomass to be predicted, historical pyrolysis parameters, historical pyrolysis products, and carbon content, hydrogen content, nitrogen content, and oxygen content corresponding to the historical characteristic parameters;

[0037] A model determination module, used to determine a first prediction model and a second prediction model with the smallest prediction error value among multiple models to be selected according to the historical data, wherein the first prediction model is obtained by training with the historical characteristic parameters and the corresponding carbon content, hydrogen content, nitrogen content and oxygen content as samples; and the second prediction model is obtained by training with the carbon content, hydrogen content, nitrogen content and oxygen content corresponding to the historical characteristic parameters, historical pyrolysis parameters and historical pyrolysis products as samples;

[0038] A second acquisition module is used to acquire current characteristic parameters and current pyrolysis parameters of the biomass to be predicted;

[0039] A first prediction module, used for inputting the current characteristic parameters into the first prediction model, and obtaining predicted carbon content, hydrogen content, nitrogen content and oxygen content through the first prediction model;

[0040] The second prediction module is used to input the predicted carbon content, hydrogen content, nitrogen content and oxygen content and current pyrolysis parameters into the second prediction model, and obtain the pyrolysis product of the biomass to be predicted through the second model.

[0041] The present invention also provides a computer device, comprising a memory, a processor and a computer program stored in the memory, wherein the processor executes the computer program to implement any one of the steps in the method for predicting biomass pyrolysis products.

[0042] The present invention also provides a computer-readable storage medium, on which a computer program is stored, characterized in that when the computer program is loaded by a processor, it can execute any one of the steps in the method for predicting biomass pyrolysis products.

[0043] The method for predicting biomass pyrolysis products provided by the present invention has the following beneficial effects: on the one hand, by using historical data of the biomass to be predicted to determine a model with the smallest prediction error value among multiple models to be selected for prediction, a prediction model with higher accuracy can be selected to improve the prediction accuracy of pyrolysis products; on the other hand, since different models have different performances in predicting carbon, hydrogen, nitrogen and oxygen content and pyrolysis products, the accuracy of pyrolysis product prediction can be further improved by predicting the carbon, hydrogen, nitrogen and oxygen content and pyrolysis products respectively using two models. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] In order to more clearly illustrate the embodiment of the present invention and its design scheme, the following briefly introduces the drawings required for this embodiment. The drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0045] Figure 1 A schematic flow chart of a method for predicting biomass pyrolysis products provided in an embodiment of the present invention;

[0046] Figure 2 A schematic flowchart of a model training provided in an embodiment of the present application;

[0047] Figure 3 A discrete graph of C, H, O, and N under different prediction results and a discrete graph of solid, liquid, and gas under different prediction results provided in an embodiment of the present application;

[0048] Figure 4 A schematic diagram of the distribution of the mean square error MSE of C, H, O, and N predicted by different models provided in an embodiment of the present application;

[0049] Figure 5A schematic diagram of the distribution of the mean square error (MSE) of solid, liquid and gas predicted by different models provided in the embodiments of the present application;

[0050] Figure 6 A histogram of biomass characteristics to be predicted provided in an embodiment of the present application;

[0051] Figure 7 A histogram of pyrolysis conditions provided in an embodiment of the present application;

[0052] Figure 8 A scatter plot of biomass characteristics provided in an embodiment of the present application;

[0053] Fig. 9 A scatter plot of pyrolysis conditions provided in an embodiment of the present application;

[0054] Fig.10 A correlation heat map provided for an embodiment of the present application;

[0055] Fig.11 A schematic diagram of a feature importance analysis provided in an embodiment of the present application;

[0056] Fig.12 A structural schematic diagram of a device for predicting biomass pyrolysis products provided in an embodiment of the present application. DETAILED DESCRIPTION

[0057] In order to enable those skilled in the art to better understand the technical solution of the present invention and implement it, the present invention is described in detail below in conjunction with the accompanying drawings and specific embodiments. The following embodiments are only used to more clearly illustrate the technical solution of the present invention, and cannot be used to limit the scope of protection of the present invention.

[0058] In the description of the present invention, the terms "first", "second", etc. are used for descriptive purposes only and are not to be understood as indicating or implying relative importance. In the description of the present invention, it should be noted that, unless otherwise clearly specified or limited, the terms "connected" and "connected" should be understood in a broad sense. For example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be directly connected or indirectly connected through an intermediate medium. For those of ordinary skill in the art, the specific meanings of the above terms in the present invention can be understood according to the specific circumstances. In the description of the present invention, unless otherwise specified, "plurality" means two or more, which will not be described in detail here.

[0059] Before introducing the execution steps of the scheme, the relevant background and related technologies of the scheme as well as the terminology involved in the embodiments of the present application are first introduced.

[0060] First, the relevant background of this plan is introduced.

[0061] Biomass pyrolysis is the process of decomposing biomass into gas, liquid and solid products by heating under anaerobic or micro-aerobic conditions. Biomass absorbs heat during the heating process and the temperature gradually rises. When the temperature reaches a certain level, the biomass begins to decompose and generate gas, liquid and solid products, which are separated and collected by corresponding collection devices.

[0062] The gaseous products of biomass pyrolysis mainly include hydrogen, methane, carbon dioxide and carbon monoxide, etc. These gases can be used for power generation and heating, and can also be used as chemical raw materials. The liquid products mainly include bio-oil, organic acids and phenols. Bio-oil has a high calorific value and can be used for combustion to generate electricity, heating, or produce biochemicals. The solid products mainly include biochar and ash. Biochar has a high porosity and specific surface area, and can be used as an adsorbent and catalyst, and is widely used in the fields of environmental protection and chemical industry.

[0063] The biomass pyrolysis process is affected by many factors. Temperature is a key factor. Increasing the temperature helps decompose biomass and improves the yield and quality of the product. The longer the reaction time, the more complete the biomass decomposition, and the higher the yield and quality of the product. The type, particle size and moisture content of the biomass also affect the pyrolysis process. Generally speaking, the smaller the particle size and the lower the moisture content, the easier it is to decompose. Different types of reactors also have a significant impact on the biomass pyrolysis process and product yield.

[0064] In recent years, biomass pyrolysis technology has received widespread attention and has made significant research progress. Researchers have deeply explored the chemical reactions in the process of biomass pyrolysis, revealed the generation rules of gas, liquid and solid products, and provided a theoretical basis for optimizing biomass pyrolysis technology. There are many ways to predict the products of biomass pyrolysis, among which the more mainstream is to experimentally predict the product composition through machine learning. Python is an efficient general-purpose programming language suitable for various data science tasks. It has a wide range of third-party library support and provides many advanced data processing, artificial intelligence, deep learning and visualization functions. In addition, Python has a large community support, easy to obtain various common and advanced artificial intelligence deep learning algorithms, and can be well integrated with other programming languages. When processing huge biomass pyrolysis prediction data, Python is more flexible and versatile, and can easily integrate other libraries and languages ​​to meet more data analysis and prediction needs.

[0065] In recent years, China has made remarkable progress in the field of biomass pyrolysis, covering technological innovation, diversified biomass utilization, application of pyrolysis products, environmental and economic benefits, etc. These research results have not only promoted the development of biomass pyrolysis technology, but also played an important role in achieving sustainable development and addressing climate change.

[0066] First, in terms of technological innovation, China has developed a variety of efficient pyrolysis reactors, such as fluidized bed reactors and fixed bed reactors. These reactors improve the efficiency of the pyrolysis process and the quality of the products. For example, a research team from Beijing University of Chemical Technology has developed a new type of fluidized bed pyrolysis reactor that significantly improves the yield and quality of bio-oil. By optimizing the pyrolysis reaction conditions, researchers are able to better control the pyrolysis process, thereby improving the purity and yield of the products.

[0067] Secondly, China's research covers the utilization of a variety of biomass materials, including agricultural waste, forestry waste and urban organic waste. Through the pyrolysis research of different biomass materials, researchers have found the optimized reaction conditions, thereby improving the utilization rate of various products. Research from East China University of Science and Technology shows that by adjusting the pyrolysis temperature and time, the pyrolysis products of rice husks and wood chips can be optimized to have higher utilization value.

[0068] In terms of the application of pyrolysis products, research on bio-oil, syngas and biochar has made significant progress. Bio-oil can be further refined into high-quality biofuels through catalytic upgrading, while biochar has also shown great potential in soil improvement and carbon sequestration. For example, research by the Chinese Academy of Sciences has shown that biochar has a significant effect in improving soil structure and fertility, helping to improve agricultural productivity and environmental sustainability.

[0069] In addition, through life cycle analysis and economic analysis, the study found that biomass pyrolysis has significant advantages in reducing greenhouse gas emissions and providing renewable energy. Related studies also pointed out that with the support of policies and technological progress, the economic efficiency of biomass pyrolysis has been significantly improved. The team of Tsinghua University proved through economic model analysis that biomass pyrolysis has a good market prospect in China and plays an important role in achieving energy structure adjustment and environmental protection goals.

[0070] At present, China has made outstanding achievements in the research and application of biomass pyrolysis. Through technological innovation, diversified biomass utilization, efficient application of pyrolysis products, and comprehensive evaluation of environmental and economic benefits, it has promoted the development of biomass pyrolysis technology and provided important support for the development of renewable energy and environmental protection. With the deepening of research and the continuous advancement of technology, biomass pyrolysis will continue to play an important role in China, helping to achieve sustainable development and respond to the challenges of global climate change. Biomass pyrolysis technology will be more widely used and developed in the future.

[0071] Similarly, international research in the field of biomass pyrolysis has also made significant progress. Researchers around the world have developed a variety of efficient pyrolysis reactors and processes. For example, the fast pyrolysis reactor at Iowa State University in the United States is able to process a variety of biomass materials and optimize temperature and atmosphere conditions to improve the yield and quality of bio-oil. In addition, the Canadian research team has made a breakthrough in low-temperature catalytic pyrolysis technology, significantly improving the yield and quality of bio-oil and reducing the content of unsaturated compounds by introducing catalysts. European research teams have extensively studied the pyrolysis process of agricultural waste, forestry waste and urban organic waste. Research by the Fraunhofer Institute in Germany shows that by optimizing pyrolysis conditions, materials such as wood chips and straw can be converted into high-value-added bio-oil and biochar, and biochar is also used for soil improvement and carbon sequestration. The Chalmers University research team in Sweden has made breakthroughs in the refining and upgrading of bio-oil, using new catalysts to significantly improve the calorific value and stability of bio-oil. The application of biochar in agriculture and environmental governance has also been widely recognized, such as its positive role in improving soil structure and increasing soil fertility. The CENER Research Center in Spain has successfully converted agricultural waste such as olive pits and grape pomace into bio-oil and biochar using biomass pyrolysis technology. These products can not only be used as energy, but also for soil improvement and carbon sequestration, demonstrating the multiple benefits of pyrolysis technology. In addition, a research team at Kyoto University in Japan has developed a new type of rapid pyrolysis technology that successfully converts biomass such as bamboo into high-purity bio-oil by precisely controlling the pyrolysis temperature and time. It has great potential in the production of fuels and chemicals.

[0072] Secondly, the relevant technologies involved in this application are introduced.

[0073] Machine learning has made significant progress in the field of biomass pyrolysis product prediction. Machine learning technology optimizes the biomass pyrolysis process and improves the accuracy and efficiency of product prediction through methods such as big data analysis, machine learning and deep learning. In China, the application of machine learning in the prediction of biomass pyrolysis products has performed well. Research institutions and enterprises have developed a variety of data-driven prediction models using machine learning technology. For example, research teams from Tsinghua University and the Chinese Academy of Sciences have established prediction models for biomass pyrolysis products through machine learning algorithms, which can accurately predict the yield and quality of bio-oil, biochar and gas. These models not only improve the accuracy of predictions, but also can be flexibly adjusted under different raw materials and process conditions. Domestic research also focuses on optimizing biomass pyrolysis processes through machine learning technology. Using deep learning algorithms, researchers can monitor and adjust pyrolysis reaction conditions in real time, optimize temperature, pressure and reaction time, and thus improve product yield and quality. For example, a research team from Zhejiang University has achieved remarkable results in optimizing rice husk pyrolysis processes using machine learning. China's machine learning technology has also been promoted in the practical application of biomass pyrolysis. Biomass pyrolysis products are widely used in energy, chemical industry, environmental protection and other fields, such as bio-oil as clean fuel and biochar for soil improvement and carbon sequestration. Many domestic companies and institutions have cooperated to promote the application of machine learning technology in the biomass pyrolysis industry chain.

[0074] Internationally, machine learning technology is also widely used in the prediction of biomass pyrolysis products. A research team in the United States has developed a prediction model based on machine learning and deep learning. For example, researchers at MIT used big data and AI algorithms to establish a prediction model for the pyrolysis products of different biomass materials, which greatly improved the prediction accuracy. Internationally, machine learning emphasizes interdisciplinary cooperation in biomass pyrolysis research. For example, a European research team combined chemical engineering, materials science and computer science to promote the application of machine learning in biomass pyrolysis through multidisciplinary integration. The research team of Imperial College London in the United Kingdom used machine learning technology to optimize the catalyst selection and process conditions for biomass pyrolysis. Some intelligent biomass pyrolysis plants have also emerged internationally, which use machine learning technology to achieve intelligent control and optimization of the entire process. These plants use AI for real-time monitoring and regulation, which improves production efficiency and product quality. For example, BASF in Germany has achieved fully automated control of biomass pyrolysis processes through machine learning technology. In terms of international cooperation, the application of AI in the field of biomass pyrolysis has also received widespread attention. International cooperation and exchanges have promoted the rapid development and application promotion of technology. For example, a European research team combined chemical engineering, materials science, and computer science to promote the application of AI in biomass pyrolysis through multidisciplinary integration. This interdisciplinary collaboration provides new perspectives and methods for biomass pyrolysis research and promotes the advancement and application of technology.

[0075] Prediction through artificial intelligence has become a mainstream trend. AI prediction mainly relies on machine learning algorithms that can learn patterns from historical data and use them to predict future trends or outcomes.

[0076] Biomass pyrolysis is a thermochemical process that converts biomass into bio-oil, gas, and solid charcoal. Predicting biomass pyrolysis products usually involves complex chemical and physical processes that are affected by multiple factors such as feedstock type, pyrolysis temperature, reaction time, heating rate, etc. Selecting a suitable machine learning algorithm for prediction requires considering the algorithm's ability to handle nonlinear relationships, the number and type of input features, and the interpretability of the algorithm. This solution first requires fixing the feature values ​​and then selecting a suitable algorithm for product prediction. The selected algorithm should have characteristics such as multivariate prediction ability, adaptability, computational efficiency, interpretability, accuracy, and data requirements.

[0077] Finally, four machine learning methods related to this application are introduced.

[0078] Random Forest: An ensemble learning method that improves the accuracy and stability of predictions by building multiple decision trees. The basic idea of ​​this method is to train multiple decision trees by randomly selecting parts of the data and features, and then combine these trees to improve the accuracy of predictions. Since each decision tree is trained by randomly selecting data and features, the random forest method can capture the diversity and complexity in the data, thereby improving the accuracy of predictions. The random forest method can reduce the risk of overfitting by randomly selecting data and features. When training each decision tree, the randomly selected data and features only account for a part of the total data and features, which can avoid the model's excessive dependence on the training data and improve the generalization ability of the model.

[0079] At the same time, the random forest method is relatively simple and easy to implement and apply. Since it is based on decision trees, the existing decision tree algorithm can be used to implement the random forest method. In addition, the random forest method can be combined with other machine learning algorithms to improve prediction performance.

[0080] The random forest method has significant advantages in the prediction of biomass pyrolysis products. First, it can process multivariate data and adapt to a variety of reaction conditions and raw material types in the biomass pyrolysis process. By constructing multiple decision trees and combining their outputs, the random forest method can capture complex nonlinear relationships and improve the accuracy of predictions. Secondly, it can effectively deal with missing data problems and improve the robustness of the model. In addition, the random forest method can also evaluate the importance of each input feature, help identify the key factors affecting product quality and yield, and thus optimize process parameters. In practical applications, researchers used the random forest method to establish a prediction model for biomass pyrolysis products, successfully predicted the yield and quality of products such as bio-oil and biochar, and optimized the pyrolysis process conditions. In the future, combined with more advanced algorithms and high-quality data, the application prospects of the random forest method in the field of biomass pyrolysis will be broader, providing important support for process optimization and industrialization.

[0081] However, the random forest method also has some disadvantages, such as high computational cost and poor model interpretability. The high computational cost is because the random forest method needs to train multiple decision trees, which consumes more computing resources. The poor model interpretability is because the random forest method is composed of multiple decision trees, and it is difficult to explain the prediction process of a single decision tree.

[0082] In general, the random forest method is an efficient, stable and flexible prediction method, which is widely used in various fields such as machine learning, data mining, financial prediction, etc. By reasonably adjusting parameters and optimizing models, the random forest method can achieve better prediction performance.

[0083] Decision tree: It is a commonly used machine learning algorithm for classification and regression analysis. Its basic idea is to represent a series of decision rules through a tree structure, where each node represents a feature, each branch represents a decision, and the final leaf node represents a category or predicted value. The learning process of a decision tree includes two main steps: tree construction and tree pruning. The tree construction process is as follows: First, select a feature and a threshold for the feature to divide the data set into two subsets; second, repeat step 1 for each subset until all data are assigned to leaf nodes or the stopping condition is met (such as reaching the preset tree depth or the node purity is high enough); finally, repeat the above process until the entire decision tree is constructed.

[0084] There are many characteristics of decision trees, including: clear structure, easy to understand and explain; the algorithm is relatively simple and easy to program without a lot of data preprocessing; decision trees can handle different types of data without complex preprocessing steps, and can be easily extended to situations involving a large number of features and samples.

[0085] The decision tree method has many practical uses in predicting biomass pyrolysis products. Decision trees have powerful feature selection and importance assessment capabilities, and can automatically select and sort the features that have the greatest impact on the prediction results, helping researchers identify key parameters (such as temperature, time, type of raw materials, etc.). In addition, decision trees can handle complex nonlinear relationships without assuming linear relationships in the data, which is particularly important for the biomass pyrolysis process. The decision tree structure is simple and intuitive, easy to interpret, and each decision path clearly shows the impact of features on the prediction results, which is convenient for researchers to understand and optimize process parameters. In addition, decision trees can handle continuous and discrete data, reduce the complexity of data preprocessing, and adapt to a variety of data types in biomass pyrolysis research (such as temperature, time, type of raw materials, chemical composition, etc.). Decision trees have good robustness to noise and outliers in the data, and can still provide reliable prediction results when there is noise and error in the data. The training and prediction process of decision trees is relatively fast, suitable for large-scale data sets, and meets the needs of rapid iterative experiments and optimization of process parameters. Decision trees can also further improve the prediction performance through ensemble learning methods (such as random forests, gradient boosting trees, etc.), combining the advantages of multiple decision trees to provide higher prediction accuracy and stability. In addition, decision trees are not only suitable for single-objective prediction, but can also be extended to multi-objective prediction. In biomass pyrolysis research, decision trees can simultaneously predict the yield and quality of multiple products and provide a comprehensive process optimization solution. In summary, the decision tree method has significant practicality in the prediction of biomass pyrolysis products due to its powerful feature selection ability, ability to handle nonlinear relationships, interpretability, flexibility in handling different types of data, robustness, fast computing ability, and multi-objective prediction.

[0086] XGBoost: is a popular machine learning library that implements the Gradient Boosting Machine (GBM) algorithm. XGBoost was created by Dr. Tianqi Chen in 2014 and quickly became a powerful and efficient tool in the field of data science and machine learning. The core idea of ​​XGBoost is to iteratively build decision trees, each of which aims to correct the error predicted by the previous tree. This method iteratively selects the optimal split point to reduce the prediction error by maximizing the gradient of the objective function. The advantage of XGBoost lies in its high efficiency and scalability, which enables efficient model training on large data sets.

[0087] XGBoost has three major characteristics: high efficiency, flexibility, and scalability. It can also reduce the risk of overfitting and improve the generalization ability of the model through regularization terms and pruning strategies, such as tree pruning and feature pruning. It also supports a variety of objective functions, including binary classification, multi-classification, regression, sorting, etc., making it suitable for various types of machine learning tasks.

[0088] Biomass pyrolysis is a thermochemical process that decomposes organic materials into gas, liquid, and solid products. The composition and yield of its products are affected by many factors, such as feedstock type, reaction temperature, and heating rate. Traditional statistical methods often seem to be unable to deal with these complex nonlinear relationships. XGBoost, as an advanced machine learning algorithm, can effectively capture these complex relationships and provide high-precision prediction results. XGBoost is based on the gradient boosting framework and builds a powerful prediction model by integrating multiple weak learners (usually decision trees). Its core advantage lies in processing high-dimensional data and preventing overfitting. In the prediction of biomass pyrolysis products, XGBoost can accurately predict the yield and composition distribution of gas, liquid, and solid products by optimizing model parameters. In addition, XGBoost also has the ability to handle missing values ​​and heterogeneous data, which is particularly important for the actual biomass pyrolysis process because experimental data are often incomplete or noisy. In practical applications, researchers can use the XGBoost model to analyze the effects of different experimental conditions on pyrolysis products and find potential optimization directions. For example, by inputting different feedstock characteristics and reaction conditions, the model can predict the best combination of products, thereby guiding experimental design and industrial applications. Compared with other machine learning methods, XGBoost has faster training speed, higher prediction accuracy, and can automatically perform feature selection, further improving the interpretability and practicality of the model.

[0089] XGBoost has performed well in various competitions and practical applications, especially in data science competitions such as Kaggle, where it is often used as a benchmark algorithm and is a key component of many winning solutions. Due to its powerful performance and flexibility, XGBoost has become an indispensable tool in the data science and machine learning communities.

[0090] GDBT: It is the abbreviation of Gradient Boosting DecisionTree, which is a variant of the Gradient Boosting Machine (GBM) algorithm. Gradient Boosting Machine is a powerful machine learning algorithm widely used in classification and regression problems. GDBT improves the performance of the model by optimizing the structure of the decision tree based on GBM. The core idea of ​​GDBT is to iteratively build decision trees, and each tree aims to correct the error predicted by the previous tree. Unlike GBM, GDBT considers not only the gradient of the objective function but also the Hessian matrix of the objective function (that is, the square of the gradient) during each tree construction process. In this way, GDBT can better capture the complex relationships in the data, thereby improving the performance of the model.

[0091] Biomass pyrolysis refers to the process of heating and decomposing biomass under anoxic conditions to produce products such as char, bio-oil, and gas. The distribution and characteristics of pyrolysis products are affected by many factors, such as biomass type, reaction temperature, and heating rate. GDBT can automatically process complex data sets and perform feature selection to extract the features that have the greatest impact on product distribution and improve the accuracy of the prediction model. Using biomass composition and reaction conditions, GDBT can effectively predict the relative proportions of char, bio-oil, and gas generated at different temperatures. Compared with traditional multivariate regression models, GDBT has more advantages in dealing with nonlinear relationships and complex interaction effects. Based on the prediction results of GDBT, pyrolysis process parameters can be optimized. By simulating the product distribution under different reaction conditions, the optimal operating conditions can be found to maximize the yield of the desired product. In addition, GDBT can interpret the model output through feature importance analysis, helping researchers understand which factors have a significant impact on the generation of pyrolysis products, thereby guiding further experimental design and process improvement. In general, GDBT has demonstrated strong practicality in the prediction and optimization of biomass pyrolysis products. Its excellent data processing capabilities, predictive performance and interpretation capabilities make it a powerful tool for studying and optimizing biomass pyrolysis processes. In the future, it is expected to play a greater role in promoting the development and utilization of renewable energy.

[0092] The following is a detailed introduction to the implementation process of this solution.

[0093] Embodiment 1:

[0094] Figure 1 A schematic flow chart of a method for predicting biomass pyrolysis products provided in an embodiment of the present application.

[0095] For example, Figure 1 As shown, the execution steps of this method are as follows:

[0096] S110, obtaining historical data of biomass to be predicted.

[0097] The biomass to be predicted is the biomass for which pyrolysis product prediction is required, and the historical data includes historical characteristic parameters of the biomass to be predicted, historical pyrolysis parameters, historical pyrolysis products, and carbon content, hydrogen content, nitrogen content, and oxygen content corresponding to the historical characteristic parameters. Optionally, relative content analysis features (e.g., cellulose Ce, hemicellulose He, and lignin Li content) may also be included.

[0098] Alternatively, historical data may be obtained via the Internet, and the biomass to be predicted may include agricultural products, forest plants, algae, and the like.

[0099] The historical characteristic parameters include the fixed carbon FC, volatile matter VM and ash Ash contents of the biomass to be predicted in the historical data, and the historical pyrolysis parameters include the final temperature FT, heating rate HR, heating time HT, sweep flow rate FR and particle size PS of the biomass to be predicted in the historical data.

[0100] S120, determining a first prediction model and a second prediction model with the smallest prediction error value among multiple models to be selected according to historical data.

[0101] Among them, the first prediction model is obtained by training with historical characteristic parameters and corresponding carbon content, hydrogen content, nitrogen content and oxygen content as samples; the second prediction model is obtained by training with carbon content, hydrogen content, nitrogen content and oxygen content corresponding to historical characteristic parameters, as well as historical pyrolysis parameters and historical pyrolysis products as samples. That is, the first prediction model is a model for predicting carbon content, hydrogen content, nitrogen content and oxygen content; the second model is a model for predicting pyrolysis products.

[0102] Therefore, the multiple models to be selected include multiple models for predicting carbon content, hydrogen content, nitrogen content and oxygen content, and multiple models for predicting pyrolysis products. The number of models to be selected is not specifically limited in the embodiment of the present application, and the types may include random forest model, decision tree model, XGBoost model and GDBT model.

[0103] A possible implementation method, the implementation process of the above S120 is as follows: inputting historical characteristic parameters into multiple models for predicting carbon content, hydrogen content, nitrogen content and oxygen content, and inputting the carbon content, hydrogen content, nitrogen content and oxygen content corresponding to the historical characteristic parameters and historical pyrolysis parameters into multiple models for predicting pyrolysis products; predicting the carbon content, hydrogen content and oxygen content corresponding to the historical characteristic parameters to obtain a first result; predicting the pyrolysis products of the biomass to be predicted to obtain a second result; comparing the first result with the carbon content, hydrogen content, nitrogen content and oxygen content corresponding to the historical characteristic parameters to determine multiple mean square errors (MSEs), and determining the prediction model corresponding to the minimum value of the multiple mean square errors (MSEs) as the first prediction model; comparing the second result with the historical pyrolysis products to determine multiple mean square errors (MSEs), and determining the prediction model corresponding to the minimum value of the multiple mean square errors (MSEs) as the second prediction model.

[0104] The first result includes the predicted carbon content, hydrogen content, nitrogen content and oxygen content output by multiple models, and the second result includes the predicted pyrolysis products output by multiple models.

[0105] Furthermore, the mean square error (MSE) of multiple models can be calculated by the following formula, and the model corresponding to the minimum mean square error value is determined as the first model and the second model. The formula is as follows:

[0106]

[0107] Where n is the number of samples, Yi is the i-th true value, and Fi is the i-th predicted value.

[0108] Mean square error (MSE) is an important indicator to measure the accuracy of the prediction model. It has three important characteristics that can help prediction: First, MSE eliminates the positive and negative signs of the errors by squaring the error values, ensuring that all errors are positive numbers. This enables MSE to more accurately reflect the deviation between the predicted values ​​and the true values. Second, MSE calculates the mean of the errors, which provides a global measure that can reflect the prediction performance of the model on the overall data. Finally, the unit of MSE is the square of the predicted value unit, which means that the value of MSE will increase with the increase of the predicted value unit.

[0109] Optionally, before making predictions through multiple models to be selected, it is necessary to first train the first type of model (model for predicting C, H, O, and N content) and the second type of model (model for predicting pyrolysis products) among the multiple models. The model training process is introduced below.

[0110] Step 1: Preprocess the historical data to obtain the first data set.

[0111] Optionally, preprocessing may include deleting missing data and duplicate data to ensure the validity of the data set.

[0112] Step 2: Divide the first dataset into a training dataset and a test dataset.

[0113] Among them, the training data set is mainly used for model training, and the test data set is mainly used for model verification.

[0114] Step 3: Train multiple models to be selected using the training data set.

[0115] A possible implementation method is to implement step 3 as follows: using the moisture content M, ash content ASH, volatile matter VM, and fixed carbon FC in the training data set as characteristic variables, and using the carbon content, hydrogen content, nitrogen content, and oxygen content corresponding to the historical characteristic parameters as target variables to train multiple models for predicting the carbon content, hydrogen content, nitrogen content, and oxygen content; using the carbon content, hydrogen content, nitrogen content, and oxygen content corresponding to the historical characteristic parameters in the training data set, as well as historical pyrolysis parameters as characteristic variables, and using the historical pyrolysis products as target variables to train the model for predicting the pyrolysis products.

[0116] Next, combine Figure 2 The specific process of model training is explained.

[0117] Embodiment 2:

[0118] Figure 2 A schematic flow chart of a model training provided in an embodiment of the present application, exemplarily, as Figure 2 As shown in the figure, the model training process is divided into the following steps:

[0119] S210, input the independent variables in the training set.

[0120] S220, calling a decision tree model.

[0121] S230, training model.

[0122] S240, use the trained model to predict the data in the test set.

[0123] S250, error analysis.

[0124] S260, give the analysis result.

[0125] Specifically, in model training, the model can be trained through two independent programs. For example, in the first program, a machine learning algorithm (e.g., random forest method) is used to predict the contents of C, H, O, and N. First, the required Python libraries are imported, including the Pandas library for data processing, the Scikit-learn library for model training, and the Joblib library for saving the model. Secondly, the data set is loaded from the specified Excel file, which contains the historical experimental data of biomass pyrolysis to be predicted. Figure 2 As shown, the data set is split into a training set and a test set for model training and evaluation. The selected features include moisture content M, ash ASH, volatile matter VM and fixed carbon FC, and the target variable is chemical composition (C, H, O, N). The decision tree regression model is trained using the training data set. The test set is predicted, and the mean square error MSE of the prediction results is calculated to evaluate the model performance.

[0126] In the second program, in order to ensure the accuracy of the training model, it is necessary to input the data of PS, FT, HR, FR and C, H, O, N in the original data set, and use the adjusted model (for example, random forest model) to obtain the predicted MSE of the solid-liquid-gas ratio.

[0127] Exemplarily, taking the program of random forest + random forest as an example, the process of training the model is introduced: This program uses the random forest regression model to predict the element content and product distribution in the process of biomass pyrolysis. First, the program loads the data file in Excel format from the specified path, and uses the moisture content M, ash ASH, volatile matter VM, and fixed carbon FC as input features, and carbon C, hydrogen H, oxygen O, and nitrogen N as target variables for model training and prediction. Then the program uses carbon C, hydrogen H, oxygen O, nitrogen N, particle size PS, pyrolysis final temperature FT, heating rate HR, and sweep rate FR as input features, and solid SOLID, liquid LIQUID, and gas GAS as target variables for training and prediction of the second model. Finally, the program outputs the mean square error MSE of the two models to evaluate the prediction accuracy of the model, thereby helping to optimize the pyrolysis process and improve product quality.

[0128] Since the training process of the two models is independent, they can be introduced through the combination of models in the following example.

[0129] Combination 1: Random Forest + Random Forest, First, introduce the training process of the first model: import the necessary libraries, including pandas for data manipulation and analysis, scikit-learn for model building and evaluation, and joblib for model serialization and saving. Read the data from the specified Excel file and store it in a pandasDataFrame named data. Then, select the feature variables (M, ASH, VM, FC) and the target variables (C, H, O, N), and then use the train_test_split function to split the data into training and test sets, where the test set accounts for 20% of the total data, and set random_state to 42 to ensure the repeatability of the results. Then, instantiate the RandomForestRegressor model rf1 and set the number of trees to 100. Use the fit method to train the model on the training data (X1_train and y1_train), and then predict on the test data (X1_test) to get the predicted value y1_pred. Use the mean_squared_error function to calculate the mean squared error (MSE) between the predicted results and the true values, and set multioutput = 'raw_values' to get a separate error value for each output variable.

[0130] Secondly, the training process of the second model is introduced: more feature variables (including C, H, O, N, PS, FT, HR, and FR) and target variables (SOLID, LIQUID, and GAS) are selected, and the train_test_split function is used again to split the data into training and test sets, with the test set accounting for 20%. Then, another RandomForestRegressor model rf2 is instantiated and wrapped into a multi-output model multi_rf2 using MultiOutputRegressor. The multi-output model is trained on the training data (X2_train and y2_train) using the fit method, and predicted on the test data (X2_test) to obtain the predicted value y2_pred. The mean_squared_error function is used to calculate the mean square error MSE between the predicted result and the true value.

[0131] Among them, train_test_split is a utility function in scikit-learn that is used to randomly split the dataset into training and test sets. The size of the test set is controlled by specifying the test_size parameter, and the random_state parameter is used to ensure the repeatability of the results. The purpose of splitting the dataset is to evaluate the performance of the model after training and prevent data leakage. Pandas is an open source library for data manipulation and analysis, especially for processing tabular data. Its main data structure is DataFrame, which allows data to be processed in a spreadsheet-like manner.

[0132] Pandas provides many data manipulation functions, such as reading and writing data in different formats (CSV, Excel, SQL, etc.), data cleaning, data transformation, and data aggregation. Scikit-learn is an important tool in the field of machine learning, providing many commonly used machine learning algorithms and evaluation methods. The RandomForestRegressor model used in this article is a regression model based on the random forest algorithm. It improves the accuracy and stability of the model by integrating the prediction results of multiple decision trees. The MultiOutputRegressor wrapper allows processing multi-output regression problems, enabling a single regression model to predict multiple target variables at the same time. The mean_squared_error function is used to evaluate the prediction performance of the model. The mean squared error MSE is a commonly used evaluation indicator in regression problems, which measures the average squared difference between the predicted value and the true value. joblib is used to serialize Python objects, which is usually used to save and load models. It is faster than the built-in pickle module, especially for objects with large amounts of data.

[0133] MultiOutputRegressor is a wrapper in scikit-learn for multi-output regression problems. It wraps a single-output regressor into a multi-output regressor, enabling it to predict multiple target variables simultaneously. The advantage of this wrapper is that it simplifies the modeling process of multi-output problems and trains a sub-model independently for each output variable.

[0134] mean_squared_error is an evaluation function in scikit-learn that calculates the mean square error (MSE) between the predicted value and the true value. MSE is a commonly used evaluation metric in regression problems that measures the average squared difference between the predicted value and the true value.

[0135] Combination 2: Decision tree + decision tree, the same as the above process, import the necessary libraries, and store the historical data in a DataFrame. Some feature variables and target variables are selected, and the data is split into training sets and test sets using the train_test_split function to ensure the fairness of model training and evaluation and avoid model performance higher than the test set accounting for 20% of the total data due to data leakage. The mean squared error MSE is calculated using the mean_squared_error function to evaluate the predictive performance of the model. In the random forest model, the selected feature variables include moisture content M, ash ASH, volatile matter VM, and fixed carbon FC, and the target variables include carbon C, hydrogen H, oxygen O, and nitrogen N. In the decision tree model, the selection of feature variables and target variables is the same as that of the random forest model.

[0136] In the training of the first model, the random forest model uses RandomForestRegressor and sets the number of trees to 100. After the model is trained, predictions are made using the test data and the mean square error (MSE) is calculated. The decision tree model uses DecisionTreeRegressor and sets the maximum depth to 10 to prevent overfitting. The training and evaluation process is the same as the random forest model.

[0137] The data preparation for the second model is similar to that of the first model, but more feature variables are selected, including C, H, O, N, particle size PS, final pyrolysis temperature FT, heating rate HR, and sweep rate FR, and the target variables include solid SOLID, liquid LIQUID, and gas GAS. When training the second model, the random forest model uses RandomForestRegressor again, and handles multi-output problems through MultiOutputRegressor. The decision tree model also uses DecisionTreeRegressor, and handles multi-output problems through MultiOutputRegressor. The training and evaluation process is similar to the first model.

[0138] By comparison, we can find differences in model selection and parameter settings. The random forest model uses RandomForestRegressor, which improves the accuracy and stability of the model by integrating the prediction results of multiple decision trees and reduces the risk of overfitting. The decision tree model uses DecisionTreeRegressor, which is easy to understand and explain, but prone to overfitting. The random forest model sets n_estimators = 100, which means that 100 trees are used to build the random forest, while the decision tree model sets max_depth = 10, which limits the maximum depth of the tree.

[0139] Although there are differences in model selection and parameter settings, pandas is used for data reading and processing, the train_test_split function is used for data segmentation, the fit method is used to train the model, the predict method is used for prediction, and mean_squared_error is used to calculate the mean square error MSE to evaluate the model performance.

[0140] The above examples show how to use different regression models to process the same dataset and compare their performance. This approach is very useful in real-world data science and machine learning projects to select the best model for a specific task.

[0141] Combination three: XGBoost+XGBoost. Similarly, XGBoost also uses pandas to operate and analyze the data. XGBoost is an efficient gradient boosting algorithm that implements parallel tree boosting. It has extremely high prediction accuracy and the ability to process structured data.

[0142] The first model is trained using XGBRegressor. Model xgb1 is instantiated and parameters are set: the number of trees is 100, the maximum depth is 10, and the learning rate is 0.1. These parameters are chosen to optimize the performance of the model. The model is trained on the training data using the fit method, and then predictions are made on the test data to obtain the predicted value y1_pred. Finally, the mean_squared_error function is used to calculate the mean squared error (MSE) between the predicted result and the true value, and the result is printed. The purpose of this step is to evaluate the predictive performance of the model by calculating the mean squared error to measure the difference between the predicted value and the true value. The data preparation for the second model is similar to the first model, but more feature variables (including C, H, O, N, PS, FT, HR, FR) and target variables (SOLID, LIQUID, GAS) are selected, and then the data is split into training and test sets using the train_test_split function. The purpose of this step is to create training and test datasets for the second model.

[0143] The second model is trained using XGBRegressor and handles multi-output problems through MultiOutputRegressor. Instantiate the model xgb2 and set the parameters the same as the first model. Use the fit method to train the model on the training data, and then predict on the test data to get the predicted value y2_pred. Finally, use the mean_squared_error function to calculate the mean square error MSE between the predicted result and the true value and print the result. The purpose of this step is to evaluate the prediction performance of the second model, especially its ability to handle multi-output regression problems.

[0144] Combination 4: GDBT+GDBT, import additional libraries:

[0145] sklearn.ensemble.GradientBoostingRegressor; GradientBoostingRegressor is a regression model in scikit-learn based on the GradientBoosting algorithm. Gradient boosting improves the prediction performance of the model by combining multiple weak learners (usually decision trees) to build a strong learner. The algorithm gradually reduces the prediction error at each step, allowing the model to better fit the data in each iteration. The main parameters include:

[0146] n_estimators: The number of weak learners (trees). Usually the more the better, but it will increase the computational complexity.

[0147] learning_rate: The learning rate controls the contribution of each weak learner to the model. Lower learning rates require more trees to converge, but generally provide better performance.

[0148] max_depth: The maximum depth of each tree, which controls the complexity of the model and prevents overfitting.

[0149] random_state: Random seed to ensure the repeatability of results.

[0150] Furthermore, after training the models for two purposes through the above-mentioned embodiments, the carbon content, hydrogen content and oxygen content corresponding to the historical characteristic parameters are predicted to obtain the first result; the pyrolysis products of the biomass to be predicted are predicted to obtain the second result. The implementation process is as follows: according to the trained model, the carbon content, hydrogen content and oxygen content corresponding to the historical characteristic parameters in the test data set are predicted to obtain the first result, and the pyrolysis products of the biomass to be predicted in the test data set are predicted to obtain the second result.

[0151] Furthermore, the mean square error of each output data in each model can be calculated based on the first result and the second result, so as to select a model with the smallest mean square error to improve the prediction accuracy of the selected model.

[0152] Example 3

[0153] For example, Figure 3 A discrete graph of C, H, O, and N under different prediction results and a discrete graph of solid, liquid, and gas under different prediction results provided in the embodiment of the present application are shown in FIG. Figure 3 In (a), red represents the mean square error MSE of C, dark green represents the mean square error MSE of H, blue represents the mean square error MSE of O, and light green represents the mean square error MSE of N; Figure 3 In Figure (b), red represents the mean square error (MSE) of solids, dark green represents the mean square error (MSE) of liquids, and blue represents the mean square error (MSE) of gases. By analyzing the discrete graph, the mean square error (MSE) of each feature can be obtained.

[0154] Table 1 is an example of the prediction mean square error (MSE) of different models for different data provided in the embodiments of the present application.

[0155] Table 1

[0156]

[0157]

[0158] For example, the distribution visualization diagram of the prediction mean square error MSE of each model for C, H, O, and N can be obtained through the data in Table 1 above, such as Figure 4 As shown, Figure 4 A schematic diagram of the distribution of the mean square error MSE of C, H, O, and N predicted by different models provided in the embodiment of the present application, through Figure 4 The prediction mean square error (MSE) of each model for C, H, O, and N can be intuitively obtained.

[0159] The data in Table 1 can also be used to obtain the distribution visualization of the prediction mean square error (MSE) of each model for solids, liquids and gases, such as Figure 5 As shown, Figure 5 A schematic diagram of the distribution of the mean square error MSE of solid, liquid and gas predicted by different models provided in the embodiment of the present application, through Figure 5 The prediction mean square error (MSE) of each model for solids, liquids and gases can be intuitively obtained.

[0160] Combined with Table 1, Figure 4 and Figure 5 It can be analyzed that for the prediction of element C, the XGBoost algorithm performs best, with an MSE value of only 0.58114064, which is nearly 0.5 less than other methods. This result shows that XGBoost has significant advantages in capturing the characteristics of element C and predicting its pyrolysis yield.

[0161] For the H element, the decision tree method is more suitable for prediction, and its MSE value is 0.05856539, which shows that the decision tree method is more effective and more accurate when processing H element data. Similarly, for the O element, the decision tree method also showed good prediction performance, and its MSE value was 0.92473701, which performed best among all the compared methods. For the N element, the decision tree is also the most suitable method, and its MSE value is 0.0802953, which proves the robustness and reliability of the decision tree in processing such data.

[0162] In summary, if you need to uniformly predict the biomass pyrolysis yield of the four elements C, H, O, and N, the decision tree method is one of the best solutions because it performs well on the three elements. However, if you only need to perform predictive analysis on the C element, the XGBoost method is obviously a better choice.

[0163] Secondly, for liquid products, the XGBoost method is more accurate in prediction, with an MSE value of 57.02029334. Although this value is relatively high, XGBoost is still the best method for liquid product prediction among all the compared methods. This may prove that the prediction of liquid products is more complicated than others, and advanced algorithms such as XGBoost are needed to capture the complex relationships in the data.

[0164] In summary, for the prediction of C, H, O and N, it is more recommended to use the decision tree method for unified prediction, especially for the element C, the XGBoost method is more suitable. For the prediction of products, it is more recommended to use the random forest method to predict solid and gas products, while for liquid products, the XGBoost method is more suitable.

[0165] In practical applications, the most appropriate model can be selected for prediction according to different prediction tasks to improve the accuracy and reliability of biomass pyrolysis product prediction.

[0166] When applying these methods, you can also consider the generalization ability of the model, that is, the model's performance on new, unseen data. To ensure that the model has good generalization ability, you can use techniques such as cross-validation to evaluate the performance of the model. In addition, you can also pay attention to the computational efficiency and interpretability of the model to ensure that the model is not only accurate, but also efficient and correct.

[0167] S130, obtaining current characteristic parameters and current pyrolysis parameters of the biomass to be predicted.

[0168] Among them, the current characteristic parameters and the current pyrolysis parameters are relevant parameters of the biomass to be predicted that has not yet undergone a pyrolysis process.

[0169] S140, inputting the current characteristic parameters into the first prediction model, and obtaining predicted carbon content, hydrogen content, nitrogen content and oxygen content through the first prediction model.

[0170] After determining a suitable first prediction model through the process of S130, the current characteristic parameters (moisture content M, ash content ASH, volatile matter VM, fixed carbon FC) are input into the first prediction model. Since the first prediction model is trained using historical characteristic parameters and corresponding carbon content, hydrogen content, nitrogen content and oxygen content as samples, when the input of the first prediction model is the current characteristic parameter, the first model can output the predicted carbon content, hydrogen content, nitrogen content and oxygen content corresponding to the characteristic parameter.

[0171] S150, inputting the predicted carbon content, hydrogen content, nitrogen content and oxygen content and the current pyrolysis parameters into the second prediction model, and obtaining the pyrolysis product of the biomass to be predicted through the second model.

[0172] After determining the appropriate second prediction model through the process of S130, the carbon content, hydrogen content, nitrogen content and oxygen content output by the first prediction model and the current pyrolysis parameters (final temperature FT, heating rate HR, sweep rate FR, particle size PS) are input into the second prediction model. Since the second prediction model is trained with the carbon content, hydrogen content, nitrogen content and oxygen content corresponding to the historical characteristic parameters, as well as the historical pyrolysis parameters and historical pyrolysis products as samples. Therefore, when the input of the second prediction model is the carbon content, hydrogen content, nitrogen content and oxygen content, as well as the current pyrolysis parameters, the second prediction model can output the corresponding pyrolysis products. Optionally, the content output by the second prediction model is the percentage of solid, liquid and gas in the pyrolysis products.

[0173] In one possible implementation, after obtaining the historical data of the biomass to be predicted, the historical data may be analyzed, and the pyrolysis process of the biomass to be predicted may be optimized according to the analysis result.

[0174] Specifically, historical data are analyzed to obtain a histogram of biomass characteristics to be predicted, a histogram of pyrolysis conditions, a scatter plot of biomass characteristics, a scatter plot of pyrolysis conditions, and a correlation heat map; the pyrolysis process of the heat medium to be predicted is optimized based on the histogram of biomass characteristics to be predicted, the histogram of pyrolysis conditions, the scatter plot of biomass characteristics, the scatter plot of pyrolysis conditions, and the correlation heat map.

[0175] Among them, the histogram of biomass characteristics is used to characterize the distribution of moisture content M, ash content ASH, volatile matter VM, fixed carbon FC, carbon content, hydrogen content, nitrogen content and oxygen content in the biomass to be predicted; the histogram of pyrolysis conditions is used to characterize the distribution of final pyrolysis temperature FT, heating rate HR, sweep flow rate FR, and particle size PS; the scatter plot of biomass characteristics is used to characterize the influence relationship between each characteristic of moisture content M, ash content ASH, volatile matter VM, fixed carbon FC, carbon content, hydrogen content, nitrogen content and oxygen content; the scatter plot of pyrolysis conditions is used to characterize the influence relationship between each value of final pyrolysis temperature FT, heating rate HR, sweep flow rate FR, and particle size PS; the correlation heat map is used to characterize the influence relationship between biomass characteristics and pyrolysis conditions.

[0176] Example 4

[0177] Figure 6 A histogram of biomass characteristics to be predicted provided in an embodiment of the present application, wherein: Figure 6 Figure a is a schematic diagram of the distribution of water content M; Figure 6 Figure b is a schematic diagram of the distribution of ash ASH;

[0178] Figure 6 Figure c is a schematic diagram of the distribution of volatile matter VM; Figure 6Figure d in the figure is a schematic diagram of the distribution of fixed carbon FC; Figure 6 Figure e in the figure is a schematic diagram of the distribution of carbon C. Figure 6 Figure f is a schematic diagram of the distribution of carbon and hydrogen H. Figure 6 The g diagram in the figure is a schematic diagram of the distribution of oxygen O. Figure 6 Figure h is a schematic diagram of the distribution of nitrogen N.

[0179] It should be understood that the histogram of biomass characteristics shows the distribution of each feature in the data set, where moisture content M, ash ASH, volatile matter VM and fixed carbon FC respectively reflect the proportion of water, inorganic matter, combustible gas and liquid and unvolatile carbon in the biomass sample; carbon C, hydrogen H, oxygen O and nitrogen N are the main elemental composition of biomass. These characteristics have an important influence on the yield and quality of bio-oil and biochar during pyrolysis. By analyzing the distribution of these characteristics, the pyrolysis process can be better optimized and the performance of pyrolysis products can be predicted.

[0180] Example 5

[0181] Figure 7 A histogram of pyrolysis conditions provided in an embodiment of the present application, wherein: Figure 7 Figure (a) is a schematic diagram of the distribution of the final pyrolysis temperature FT; Figure 7 Figure (b) is a schematic diagram of the distribution of heating rate HR; Figure 7 Figure (c) is a schematic diagram of the distribution of the sweep flow rate FR; Figure 7 Figure (d) is a schematic diagram of the distribution of particle size PS.

[0182] It should be understood that the histogram of pyrolysis conditions can show the distribution of the final pyrolysis temperature FT, heating rate HR, sweep rate FR, and particle size PS. The histogram of pyrolysis conditions shows the distribution of the final pyrolysis temperature, heating rate, inert gas sweep rate, and particle size, which have an important influence on the pyrolysis process and the yield and quality of its products. Analyzing these distributions helps to optimize the pyrolysis process and improve the pyrolysis efficiency and product quality.

[0183] Example 6

[0184] Figure 8 A scatter plot of biomass characteristics is provided in the embodiments of the present application. It should be understood that the scatter plot of biomass characteristics shows the relationship between different biomass characteristics, reveals how each characteristic affects each other, and provides data support for optimizing the pyrolysis process and improving product quality.

[0185] Example 7

[0186] Fig. 9A scatter plot of pyrolysis conditions is provided in an embodiment of the present application. It should be understood that the scatter plot of pyrolysis conditions reveals the relationship between the final pyrolysis temperature, heating rate, sweep rate and particle size. The scatter plot of pyrolysis conditions can provide support for understanding how each condition jointly affects the pyrolysis process and products, thereby optimizing the pyrolysis process.

[0187] Example 8

[0188] Fig.10 A correlation heat map provided for the embodiment of the present application, it should be understood that the correlation heat map shows the relationship between biomass characteristics and pyrolysis conditions, revealing how these characteristics influence each other and work together. The correlation coefficient in the figure ranges from -1 to 1, positive correlation indicates a positive relationship between characteristics, negative correlation indicates a negative relationship, and a coefficient close to 0 indicates no correlation. The heat map intuitively shows the correlation between biomass characteristics such as moisture content M, ash ASH, volatiles VM, fixed carbon FC, carbon C, hydrogen H, oxygen O, nitrogen N, etc. For example, the positive correlation between volatiles and carbon indicates that the carbon content in samples with high volatile content is also high. The correlation between pyrolysis conditions such as pyrolysis final pyrolysis temperature FT, heating rate HR, sweep rate FR, particle size PS, etc. is also revealed to help understand how these conditions influence each other. The correlation between biomass characteristics and pyrolysis conditions, such as the positive correlation between fixed carbon and pyrolysis temperature, shows that samples with high fixed carbon content have better pyrolysis effects at high temperatures.

[0189] Therefore, by analyzing the mutual influence relationship between the above factors, the pyrolysis process can be better optimized to improve the pyrolysis efficiency and the quality of the pyrolysis products.

[0190] Example 9

[0191] Optionally, you can also get a visualization of the feature importance analysis of M, Ash, VM, FC for predicting C, H, O, N through historical data, and get the importance scores of the four feature values. Fig.11 As shown, Fig.11 A schematic diagram of a feature importance analysis provided in an embodiment of the present application.

[0192] For example, according to Fig.11 Get the importance scores of the four eigenvalues:

[0193] Volatile matter VM: 0.454980; ash ASH: 0.354978; water content M: 0.128423; fixed carbon FC: 0.061619.

[0194] Fig.12 A structural schematic diagram of a device for predicting biomass pyrolysis products provided in an embodiment of the present application.

[0195] Exemplarily, the prediction device of biomass pyrolysis products includes:

[0196] The first acquisition module 121 is used to acquire historical data of the biomass to be predicted; wherein the historical data includes historical characteristic parameters of the biomass to be predicted, historical pyrolysis parameters, historical pyrolysis products, and carbon content, hydrogen content, nitrogen content, and oxygen content corresponding to the historical characteristic parameters;

[0197] A model determination module 122 is used to determine a first prediction model and a second prediction model with the smallest prediction error value among multiple models to be selected according to historical data, wherein the first prediction model is obtained by training with historical characteristic parameters and corresponding carbon content, hydrogen content, nitrogen content and oxygen content as samples; and the second prediction model is obtained by training with carbon content, hydrogen content, nitrogen content and oxygen content corresponding to the historical characteristic parameters, historical pyrolysis parameters and historical pyrolysis products as samples;

[0198] The second acquisition module 123 is used to acquire the current characteristic parameters and current pyrolysis parameters of the biomass to be predicted;

[0199] A first prediction module 124, used to input the current characteristic parameters into a first prediction model, and obtain predicted carbon content, hydrogen content, nitrogen content and oxygen content through the first prediction model;

[0200] The second prediction module 125 is used to input the predicted carbon content, hydrogen content, nitrogen content and oxygen content and the current pyrolysis parameters into the second prediction model, and obtain the pyrolysis product of the biomass to be predicted through the second model.

[0201] Each module in the above-mentioned prediction device of biomass pyrolysis products can be implemented in whole or in part by software, hardware and a combination thereof. Each of the above-mentioned modules can be embedded in or independent of a processor in a computer device in the form of hardware, or can be stored in a memory in a computer device in the form of software, so that the processor can call and execute the operations corresponding to each of the above modules.

[0202] The present invention also provides a computer device, including a memory, a processor, and a computer program stored in the memory, wherein the processor executes the computer program to implement the steps in the embodiment of the method for predicting biomass pyrolysis products. The specific implementation method can be found in the method embodiment, which will not be repeated here.

[0203] Furthermore, the present invention also provides a non-temporary computer-readable storage medium containing instructions, and a computer program is stored on the storage medium. For example, a memory containing instructions, the above instructions can be executed by a processor of a computer device to complete the above method. For example, a non-temporary computer-readable storage medium can be a ROM, a random access memory (RAM), a CD-ROM, a magnetic tape, a floppy disk, and an optical data storage device. When the computer program is executed by the processor, the steps in the embodiment of the method for predicting biomass pyrolysis products can be implemented. The specific implementation method can be found in the method embodiment, which will not be repeated here.

[0204] It will be appreciated by those skilled in the art that embodiments of the present invention may provide methods, systems or computer program products. Therefore, the present invention may take the form of a complete hardware embodiment, a complete software embodiment or an embodiment combining software and hardware. Moreover, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0205] The present invention is described with reference to flowcharts and / or block diagrams of methods, apparatus (systems) and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as a combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0206] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0207] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1A step that specifies a function in one or more boxes.

[0208] It should be pointed out that the above specific implementation methods can enable those skilled in the art to understand the invention more comprehensively, but do not limit the invention in any way. Therefore, although the present specification and embodiments have described the invention in detail, those skilled in the art should understand that the invention can still be modified or replaced by equivalents; and all technical solutions and improvements that do not deviate from the spirit and scope of the invention are included in the protection scope of the patent for the invention. Any figure mark in the claims should not be regarded as limiting the claims involved. Any simple change or equivalent replacement of the technical solution that can be obviously obtained by any technician familiar with the field within the technical scope disclosed in the present invention belongs to the protection scope of the present invention.

Claims

1. A method for predicting biomass pyrolysis products, characterized in that: include: Acquire historical data of the biomass to be predicted; wherein the historical data includes historical characteristic parameters of the biomass to be predicted, historical pyrolysis parameters, historical pyrolysis products, and carbon content, hydrogen content, nitrogen content, and oxygen content corresponding to the historical characteristic parameters; Determine a first prediction model and a second prediction model with the smallest prediction error value from a plurality of models to be selected according to the historical data; wherein the first prediction model is obtained by training with the historical characteristic parameters and the corresponding carbon content, hydrogen content, nitrogen content and oxygen content as samples; and the second prediction model is obtained by training with the carbon content, hydrogen content, nitrogen content and oxygen content corresponding to the historical characteristic parameters, historical pyrolysis parameters and historical pyrolysis products as samples; Acquiring current characteristic parameters and current pyrolysis parameters of the biomass to be predicted; Inputting the current characteristic parameters into the first prediction model, and obtaining predicted carbon content, hydrogen content, nitrogen content and oxygen content through the first prediction model; The predicted carbon content, hydrogen content, nitrogen content and oxygen content and current pyrolysis parameters are input into the second prediction model, and the pyrolysis product of the biomass to be predicted is obtained through the second model.

2. The method according to claim 1, characterized in that The multiple models to be selected include multiple models for predicting carbon content, hydrogen content, nitrogen content and oxygen content, and multiple models for predicting pyrolysis products, and determining a first prediction model and a second prediction model with the smallest prediction error value among the multiple models to be selected according to the historical data, including: Inputting the historical characteristic parameters into the multiple models for predicting carbon content, hydrogen content, nitrogen content and oxygen content, and inputting the carbon content, hydrogen content, nitrogen content and oxygen content corresponding to the historical characteristic parameters and the historical pyrolysis parameters into the multiple models for predicting pyrolysis products; Predicting the carbon content, hydrogen content and oxygen content corresponding to the historical characteristic parameters to obtain a first result; predicting the pyrolysis product of the biomass to be predicted to obtain a second result; Compare the first result with the carbon content, hydrogen content, nitrogen content and oxygen content corresponding to the historical characteristic parameters to determine multiple mean square errors (MSEs), and determine the prediction model corresponding to the minimum value among the multiple mean square errors (MSEs) as the first prediction model; The second result is compared with historical pyrolysis products to determine multiple mean square errors (MSEs), and the prediction model corresponding to the minimum value of the multiple mean square errors (MSEs) is determined as the second prediction model.

3. The method according to claim 2, characterized in that Before inputting the historical characteristic parameters into the multiple models for predicting the carbon content, hydrogen content, nitrogen content and oxygen content, and inputting the carbon content, hydrogen content, nitrogen content and oxygen content corresponding to the historical characteristic parameters and the historical pyrolysis parameters into the multiple models for predicting pyrolysis products, the method further includes: Performing deduplication preprocessing on the historical data to obtain a first data set; Dividing the first data set into a training data set and a test data set; Training the plurality of models to be selected by using the training data set; The carbon content, hydrogen content and oxygen content corresponding to the historical characteristic parameters are predicted to obtain a first result; and the pyrolysis products of the biomass to be predicted are predicted to obtain a second result, including: predicting the carbon content, hydrogen content and oxygen content corresponding to the historical characteristic parameters in the test data set according to the trained model to obtain the first result, and predicting the pyrolysis products of the biomass to be predicted in the test data set to obtain the second result.

4. The method according to claim 3, characterized in that The historical characteristic parameters include water content M, ash content ASH, volatile matter VM, and fixed carbon FC. The training set is used to train the multiple models to be selected, including: The moisture content M, ash content ASH, volatile matter VM, and fixed carbon FC in the training data set are used as characteristic variables, and the carbon content, hydrogen content, nitrogen content, and oxygen content corresponding to the historical characteristic parameters are used as target variables, and a plurality of models for predicting the carbon content, hydrogen content, nitrogen content, and oxygen content are trained; The carbon content, hydrogen content, nitrogen content and oxygen content corresponding to the historical characteristic parameters in the training data set, and the historical pyrolysis parameters are used as characteristic variables, and the historical pyrolysis products are used as target variables to train multiple models for predicting pyrolysis products.

5. The method according to any one of claims 2 to 4, characterized in that: The formula for determining the mean square error MSE is: Where n is the number of samples, Yi is the i-th true value, and Fi is the i-th predicted value.

6. The method according to any one of claims 1 to 4, characterized in that The pyrolysis parameters include pyrolysis final temperature FT, heating rate HR, sweep rate FR, and particle size PS, and the method further includes: Analyzing the historical data to obtain a histogram of the biomass characteristics to be predicted, a histogram of the pyrolysis conditions, a scatter plot of the biomass characteristics, a scatter plot of the pyrolysis conditions, and a correlation heat map; Wherein, the histogram of the biomass characteristics is used to characterize the distribution of the moisture content M, ash content ASH, volatile matter VM, fixed carbon FC, carbon content, hydrogen content, nitrogen content and oxygen content in the biomass to be predicted; the histogram of the pyrolysis conditions is used to characterize the distribution of the final pyrolysis temperature FT, heating rate HR, sweep flow rate FR, and particle size PS; the scatter plot of the biomass characteristics is used to characterize the influence relationship between each of the moisture content M, ash content ASH, volatile matter VM, fixed carbon FC, carbon content, hydrogen content, nitrogen content and oxygen content; the scatter plot of the pyrolysis conditions is used to characterize the influence relationship between each value in the final pyrolysis temperature FT, heating rate HR, sweep flow rate FR, and particle size PS; the correlation heat map is used to characterize the influence relationship between the biomass characteristics and the pyrolysis conditions; The pyrolysis process of the biomass to be predicted is optimized according to the histogram of the biomass characteristics to be predicted, the histogram of the pyrolysis conditions, the scatter plot of the biomass characteristics, the scatter plot of the pyrolysis conditions and the correlation heat map.

7. The method according to any one of claims 1 to 4, characterized in that The multiple models to be selected include a random forest model, a decision tree model, an XGBoost model and a GDBT model.

8. A device for predicting biomass pyrolysis products, characterized in that: include: A first acquisition module is used to acquire historical data of the biomass to be predicted; wherein the historical data includes historical characteristic parameters of the biomass to be predicted, historical pyrolysis parameters, historical pyrolysis products, and carbon content, hydrogen content, nitrogen content, and oxygen content corresponding to the historical characteristic parameters; A model determination module, used to determine a first prediction model and a second prediction model with the smallest prediction error value among multiple models to be selected according to the historical data, wherein the first prediction model is obtained by training with the historical characteristic parameters and the corresponding carbon content, hydrogen content, nitrogen content and oxygen content as samples; and the second prediction model is obtained by training with the carbon content, hydrogen content, nitrogen content and oxygen content corresponding to the historical characteristic parameters, historical pyrolysis parameters and historical pyrolysis products as samples; A second acquisition module is used to acquire current characteristic parameters and current pyrolysis parameters of the biomass to be predicted; A first prediction module, used for inputting the current characteristic parameters into the first prediction model, and obtaining predicted carbon content, hydrogen content, nitrogen content and oxygen content through the first prediction model; The second prediction module is used to input the predicted carbon content, hydrogen content, nitrogen content and oxygen content and current pyrolysis parameters into the second prediction model, and obtain the pyrolysis product of the biomass to be predicted through the second model.

9. A computer device comprising a memory, a processor and a computer program stored in the memory, characterized in that: The processor executes the computer program to implement the steps of the method according to any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is loaded into a processor, the computer program can execute the steps of the method according to any one of claims 1 to 7.

Images