Flash BMP®: Calibration for the biochemical methane potential (BMP) of solid organic waste using NIRS

Anaerobic digestion is more and more used as renewable energy sources. It has the double advantages to reduce the waste quantities but also to produce renewable energy as electricity and/or heat using biogas. Anaerobic digestion as an industrial process has to be optimized to increase the production of methane. One method to improve the optimization is to characterize and control the substrates which are fed in the digester.
The Biochemical Methane Potential (BMP) is a commonly used parameter for substrate characterization before to use them in industrial anaerobic digester. It determines the quantity of methane (ml CH4.g-1 Volatile Solid) a substrate can potentially produce in anaerobic conditions. Thus, it is used to know if the feedstock does not change, or to optimize the co-digestion. More than 30 days are necessary to obtain the BMP value. As a consequence, this test is not suitable as a management tool at industrial scale for anaerobic digestion optimization.
Our proposal is to use near infrared spectroscopy (NIRS) to analyze the organic matter of several kinds of substrates and thus to predict the methane potential faster than the biochemical approaches.
The novelty statement of this study is the use of a wide range of substrates in a unique and global PLS model. More than 600 substrates as agro-industrial waste, green waste, energy crops, municipal solid waste, sludges and digestates were used together leading to build a robust and accurate model.
Samples were frozen and dried to allow their easy grinding (between 1 and 2 mm) which improved the homogeneity of the final sample.
The BMP values were determined in duplicate for each sample using the method described in detail in Angelidaki et al., (2009).
Spectra were measured with a BUCHI NIRFlex N-500 (Buchi, Switzerland), with Petri Dish add-on, allowing the direct analysis of an important and representative quantity of waste.
Due to the big data set, it has been decided to use calibration/validation/test sets, 50/25/25% respectively.
Different pretreatments were tested in order to reduce scattering effect: Standard Normal Variate (SNV), first and second Savitzky-Golay derivatives.
Chemometrics methods were carried out using Matlab (The Mathworks, Inc, MA).
The best model gave a RMSEP and a R² ranges of 40-60 ml CH4.g-1 VS and 0.7-0.8 respectively, depending on the kind of substrate considered. The Flash BMP® performance is therefore similar to the reproducibility of the reference BMP test.
The advantages of such a calibration are numerous:
- Fast measurement, only a couple of minutes for the spectral measurement and less than two days for sample preparation (freeze-drying and grinding)
- Larger sampling than the reference method, more than 10 g of dry matter against 1g with the reference method
- No costs for solvents, reagents, pipettes, consumables, …
- The validation is made on a large database which makes the model robust in respect of waste variability
- The possibility to measure other parameters simultaneously with the same sample with the same instrument (ex: organic matter characterization).