Rapid detection of volatile fatty acids in biogas slurry using near-infrared spectroscopy combined with optimized wavelength selection and partial least squares regression
Anaerobic fermentation is a critical technology for the resource utilization of agricultural and livestock waste, with its efficiency and stable operation heavily reliant on the dynamic changes in volatile fatty acid concentrations. However, traditional gas chromatography methods are time-consuming...
Saved in:
| Main Authors: | , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Elsevier
2025-07-01
|
| Series: | Results in Chemistry |
| Subjects: | |
| Online Access: | http://www.sciencedirect.com/science/article/pii/S221171562500445X |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | Anaerobic fermentation is a critical technology for the resource utilization of agricultural and livestock waste, with its efficiency and stable operation heavily reliant on the dynamic changes in volatile fatty acid concentrations. However, traditional gas chromatography methods are time-consuming and costly, rendering them unsuitable for rapid detection. This study employs near-infrared spectroscopy combined with partial least squares regression (PLSR) and proposes a multi-step optimization strategy to develop rapid detection models for acetic acid, propionic acid, and total acid content. Using corn straw and livestock manure as substrates, 150 biogas slurry samples were collected. Characteristic wavelengths were selected through competitive adaptive reweighted sampling and hierarchical clustering, reducing the full spectrum by approximately 85 %–90 %. Bayesian optimization was applied to simultaneously tune the characteristic wavelengths and PLSR latent variables. After optimization, the number of characteristic wavelengths for acetic acid, propionic acid, and total acid was 75, 93, and 82, respectively, with corresponding latent variables of 18, 19, and 10. Model validation yielded determination coefficients of 0.9882, 0.9452, and 0.9483, and root mean square errors of prediction of 0.1193, 0.1178, and 0.5492 for acetic acid, propionic acid, and total acid, respectively, representing reductions of approximately 40 %, 18 %, and 30 % compared to full-spectrum modeling. This method significantly enhances detection accuracy and stability, providing an efficient strategy for optimizing anaerobic fermentation processes. |
|---|---|
| ISSN: | 2211-7156 |