{"title":"用正则化的高阶多项式回归预测鼓泡流化床生物质气化产物:一种简单但高效的策略。","authors":"Michael Binns","doi":"10.1016/j.biortech.2025.133109","DOIUrl":null,"url":null,"abstract":"<p><p>For the design of biomass gasification it is beneficial to have models which can predict the composition of gas products for a wide range of different biomass feedstocks. Complex machine learning models (e.g. neural networks and tree-based methods) are now being used for this purpose which are difficult to reproduce with large numbers of parameters involved. In this study the potential for higher order polynomials is investigated for the modelling of bubbling fluidised bed gasification. To reduce the number of parameters and to avoid over-fitting Least Absolute Shrinkage and Selection Operator (LASSO) regularisation is used. This is a novel application of high order polynomial regression with regularisation which allows the prediction of hydrogen composition with coefficient of performance of 0.9228 and only 85 fitted parameters. The next best existing methods give coefficients of performance of 0.8823 and 0.868 but require 261 parameters and more than 1000 parameters respectively. So this polynomial approach is shown to give accurate model prediction with simpler model equations.</p>","PeriodicalId":258,"journal":{"name":"Bioresource Technology","volume":" ","pages":"133109"},"PeriodicalIF":9.0000,"publicationDate":"2025-12-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"0","resultStr":"{\"title\":\"Predicting biomass gasification products for bubbling fluidised beds using high order polynomial regression with regularisation: a simple but highly effective strategy.\",\"authors\":\"Michael Binns\",\"doi\":\"10.1016/j.biortech.2025.133109\",\"DOIUrl\":null,\"url\":null,\"abstract\":\"<p><p>For the design of biomass gasification it is beneficial to have models which can predict the composition of gas products for a wide range of different biomass feedstocks. Complex machine learning models (e.g. neural networks and tree-based methods) are now being used for this purpose which are difficult to reproduce with large numbers of parameters involved. In this study the potential for higher order polynomials is investigated for the modelling of bubbling fluidised bed gasification. To reduce the number of parameters and to avoid over-fitting Least Absolute Shrinkage and Selection Operator (LASSO) regularisation is used. This is a novel application of high order polynomial regression with regularisation which allows the prediction of hydrogen composition with coefficient of performance of 0.9228 and only 85 fitted parameters. The next best existing methods give coefficients of performance of 0.8823 and 0.868 but require 261 parameters and more than 1000 parameters respectively. So this polynomial approach is shown to give accurate model prediction with simpler model equations.</p>\",\"PeriodicalId\":258,\"journal\":{\"name\":\"Bioresource Technology\",\"volume\":\" \",\"pages\":\"133109\"},\"PeriodicalIF\":9.0000,\"publicationDate\":\"2025-12-01\",\"publicationTypes\":\"Journal Article\",\"fieldsOfStudy\":null,\"isOpenAccess\":false,\"openAccessPdf\":\"\",\"citationCount\":\"0\",\"resultStr\":null,\"platform\":\"Semanticscholar\",\"paperid\":null,\"PeriodicalName\":\"Bioresource Technology\",\"FirstCategoryId\":\"5\",\"ListUrlMain\":\"https://doi.org/10.1016/j.biortech.2025.133109\",\"RegionNum\":1,\"RegionCategory\":\"环境科学与生态学\",\"ArticlePicture\":[],\"TitleCN\":null,\"AbstractTextCN\":null,\"PMCID\":null,\"EPubDate\":\"2025/8/6 0:00:00\",\"PubModel\":\"Epub\",\"JCR\":\"Q1\",\"JCRName\":\"AGRICULTURAL ENGINEERING\",\"Score\":null,\"Total\":0}","platform":"Semanticscholar","paperid":null,"PeriodicalName":"Bioresource Technology","FirstCategoryId":"5","ListUrlMain":"https://doi.org/10.1016/j.biortech.2025.133109","RegionNum":1,"RegionCategory":"环境科学与生态学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"2025/8/6 0:00:00","PubModel":"Epub","JCR":"Q1","JCRName":"AGRICULTURAL ENGINEERING","Score":null,"Total":0}
Predicting biomass gasification products for bubbling fluidised beds using high order polynomial regression with regularisation: a simple but highly effective strategy.
For the design of biomass gasification it is beneficial to have models which can predict the composition of gas products for a wide range of different biomass feedstocks. Complex machine learning models (e.g. neural networks and tree-based methods) are now being used for this purpose which are difficult to reproduce with large numbers of parameters involved. In this study the potential for higher order polynomials is investigated for the modelling of bubbling fluidised bed gasification. To reduce the number of parameters and to avoid over-fitting Least Absolute Shrinkage and Selection Operator (LASSO) regularisation is used. This is a novel application of high order polynomial regression with regularisation which allows the prediction of hydrogen composition with coefficient of performance of 0.9228 and only 85 fitted parameters. The next best existing methods give coefficients of performance of 0.8823 and 0.868 but require 261 parameters and more than 1000 parameters respectively. So this polynomial approach is shown to give accurate model prediction with simpler model equations.
期刊介绍:
Bioresource Technology publishes original articles, review articles, case studies, and short communications covering the fundamentals, applications, and management of bioresource technology. The journal seeks to advance and disseminate knowledge across various areas related to biomass, biological waste treatment, bioenergy, biotransformations, bioresource systems analysis, and associated conversion or production technologies.
Topics include:
• Biofuels: liquid and gaseous biofuels production, modeling and economics
• Bioprocesses and bioproducts: biocatalysis and fermentations
• Biomass and feedstocks utilization: bioconversion of agro-industrial residues
• Environmental protection: biological waste treatment
• Thermochemical conversion of biomass: combustion, pyrolysis, gasification, catalysis.