Near Infrared Spectroscopy For Rapid Determination Of Biochemical Methane Potential Of Organic Waste

Biomass energy is the fourth largest energy source in the world and has become one of the most promising renewable energy sources.Among the existing method of treating organic waste to biomass energy,anaerobic digestion technology is one of the most widely used technologies.In the anaerobic digestion process,the biochemical methane potential?BMP expressed as methane production per kg of organic matter?is the most relevant index that is used to characterize the biogas production potential.Traditional anaerobic digestion assay for determining the biochemical methane potential of materials is time-consuming and labor-intensive.Automated measuring instruments is also expensive.Therefore,it is necessary to develop a simple,fast and reliable method to predict reliably the BMP.This paper discusses building a method for BMP prediction of organic waste based on near-infrared spectroscopy combined with Chemometric and a low-cost automated anaerobic digestion platform.The first chapter introduces the research background and significance of the paper,summarizes the research status about the BMP determination method and application of BMP experimental device.And then,the main research contents and technical routes of the subject are introduced in detail.The second chapter introduces the design and test of an automatic measuring apparatus for BMP.In a small automatic digester,the methane yield was obtained by a weight sensor that measured the weight of liquid displaced by the gas in a previously calibrated cylinder.This system consists of NI acquisition card,temperature sensor,pH sensor,weight sensor,electromagnetic pinch valve and micro peristaltic pump.LabVIEW is used to write host program to monitor gas production,temperature and pH in real time.The automatically water displacement system is added to realize the automatic monitoring function.The third chapter introduces the theoretical and steps of BMP measurement and near infrared spectra data acquiring.In total,66 different types of organic waste were collected and the BMP was determined in an naerobic batch assay performed in a self-made small digester.The results showed that the fruits had the largest BMP,the vegetables listed the second,and urban whithered leaves had the smallest one.Spectral data of the samples were acquired using a Fourier near-infrared spectrometer for subsequent processing modeling.The fourth chapter introduces the Near infrared spectra data preprocessing and building models.Firstly,the spectra data were preprocessed by Savitzky-Golay combined with multiple scattering correction?SG-MSC?method.The predicted models were developed by PLS regression and partial least squares&support vector machine based on the the full spectra data that is pre-processed with SG and MSC algorithm.The results showed that the PLS-SVM model gave better prediction,RMSECV is 74mL CH₄/g VS,RMSEP is 44mL CH₄/g VS,RPD is 2.45,RPD is0.86.The fifth chapter introduces the research on wavelength selection.In order to improve the prediction accuracy of the models and to solve the problems of large scale of spectral data,overlapping spectral information,and complicated calculation,three kinds of algorithm,genetic algorithm?GA?,ant colony algorithm?ACO?,variable importance for projection combined with ant colony algorithm?VIP-ACO?,were applied to select characteristic wavelengths.VIP-ACO coupled with PLS-SVM gave the most satisfactory results:RMSECV is 48mL CH₄/g VS,RMSEP is 29mL CH₄/g VS,R²is 0.91 and RPD is 2.92.The sixth chapter summarizes the main content of the paper and points out the inadequacy of the research work and the possible future research direction.