Biorefinery approaches are suitable for agro-residues valorization. In this work, the integration of solid-liquid extraction followed by anaerobic digestion is explored. This study aims to assess the near-infrared (NIR) spectroscopy and multivariate regression models as reliable methods to predict the methane yield of raw and extracted agro-residues. Tomato residues (ripe semi-rotten tomato-RT, green (unripe) fruit-GT, tomato plant-TB) and grape pomace residues (GP) were extracted for phenolic compounds recovery. GP is the richest substrate in phenolic compounds (55.8 mg g⁻¹, expressed gallic acid equivalents on a dry extract basis). The experimental values of biochemical methane potential (BMP) varied in the range128–307 NmL CH₄ g⁻¹, in volatile solid basis for tomato residues, while 115–177 NmL CH₄ g⁻¹ were determined for GP. The experimental BMP observed before and after extraction was statistically similar. The prediction based on the NIR-based model also exposes the same trend and shows a reasonable prediction error compared to other models. In conclusion, NIR spectroscopy and some multivariate regression models may be used for BMP prediction in a biorefinery context.

生物精炼方法适用于农业残留物的增值。在这项工作中，探索了固液萃取后厌氧消化的整合。本研究旨在评估近红外 (NIR) 光谱和多元回归模型作为预测原始和提取的农业残留物甲烷产量的可靠方法。提取番茄残渣（成熟半烂番茄-RT、绿色（未成熟）果实-GT、番茄植株-TB）和葡萄渣残渣（GP）以回收酚类化合物。GP 是酚类化合物中最丰富的底物（55.8 mg g ^-1，以干提取物为基础表示的没食子酸当量）。生化甲烷势 (BMP) 的实验值在 128–307 NmL CH ₄ g ^{-1 范围内变化}, 以挥发性固体为基础的番茄残留物，而 115–177 NmL CH ₄ g ^-1被测定为 GP。提取前后观察到的实验 BMP 在统计上是相似的。与其他模型相比，基于 NIR 模型的预测也暴露出相同的趋势，并显示出合理的预测误差。总之，近红外光谱和一些多元回归模型可用于生物精炼环境中的 BMP 预测。