Method developMent and research equipMent 201 Online measurements of volatile fatty acids in renewable raw material biogas plants by NIRS

Andrea Stockl, Hans Oechsner, Jungbluth, Thomas
An online measurement system based on near-infrared-reflectance spectroscopy shall give information on substrate condition or the fermentation progress. Hereby, NIR sensors are calibrated according to substrate-specific parameters (volatile fatty acids, like acetic acid and propionic acid) through which the process stability in the digester can be monitored online. An appropriate positioning of the NIR-sensor is very important. Two sensors were directly installed in a 400 L digester, whereas a
more » ... hird sensor was placed in a bypass. For calibration of the measurement system acetate were applied to the digester to increase the acidity artificially. The developed calibration models with "support vector regression" shows the excellence of the bypass. The statistical value of the RPD (ratio of standard deviation and standard error of prediction) for acetic acid could be increased from 1.8-2.2 in the digester to 3.3 in the bypass system. Abstract landtechnik 66 (2011), no. 3, pp. 201-204, 5 figures, 4 references n Near-infrared-reflectance spectroscopy (NIRS) is meantime applied in many different areas of agriculture and represents a rapid and non-destructive method of determining substrate-specific characteristics of samples utilising the physical-optical characteristics of the substrate. The amount of absorption, depending on the reflection of the substrate-specific contents, allows direct conclusions on the concentration of the investigated parameter. A NIRS calculation model is achieved over a statistical analysis calculated from laboratory results in combination with parallel-recorded spectra (figure 1). For this, content of organic fatty acids such as acetic acid, bu-tyric acid, propionic acid and isopropionic acid, as well as valeric and isovaleric acids that correlate with the recorded spectra, are determined. With the help of a multivariate data analysis via "support vector regression" [1] calibration models are developed and tested for the various parameters. The best models were to be later used for the estimation of volatile fatty acid concentrations in unidentified samples based on the online spectra. Model reliability and precision decide upon the quality of the estimates. To achieve a good calibration model a high variation in the spectra has to be achieved. This poses the question to what extent the position of the near-infrared sensor, and the flow speed of the substrate in front of the sensor, influences the quality of the recorded spectra and the associated calibration results. Materials and method To determine the most suitable position for the sensor in the di-gester, three NIR sensors were fitted simultaneously in a ther-mophilic 400 litre experimental biogas digester. One sensor was separately positioned in a bypass pipeline. To achieve a uniform flow of fermentation substrate towards this sensor it was positioned in a bend on the pipeline. Because rising accumulations of gas bubbles in the pipeline system could distort the NIR sensor image, the substrate was pumped upwards in the direction of the sensor. With the eccentric screw pump moving approx. 800 l/hour this meant the total digester contents flowed twice per hour past the sensor, permitting numerous varying spectra to be recorded at this point. Two further sensors were positioned at the front of the di-gester in direct contact with the digester substrate (figure 2). The substrate samples were taken via sampling-tap positioned beside the front-mounted sensors. In a preliminary trial (not included here) it could be shown that the sampling location can be selected with some flexibility in this size of digester because the agitator within the digester enables an optimum homologa-tion of the entire contents.