The world energy consumption has increased by + 195% since 1970 with more than 80% of the energy mix originating from fossil fuels, thus leading to pollution and global warming. Alternatively, pyrolysis of modern biomass is considered carbon neutral and produces value-added biogas, bio-oils, and biochar, yet actual pyrolysis processes are not fully optimized. Here, we review the use of machine learning to improve the pyrolysis of lignocellulosic biomass, with emphasis on machine learning algorithms and prediction of product characteristics. Algorithms comprise regression analysis, artificial neural networks, decision trees, and the support vector machine. Machine learning allows for the prediction of yield, quality, surface area, reaction kinetics, techno-economics, and lifecycle assessment of biogas, bio-oil, and biochar. The robustness of machine learning techniques and engineering applications are discussed.

自 1970 年以来，世界能源消耗增加了 195% 以上，其中 80% 以上的能源结构来自化石燃料，从而导致污染和全球变暖。或者，现代生物质的热解被认为是碳中性的，并产生增值沼气、生物油和生物炭，但实际的热解过程并未完全优化。在这里，我们回顾了利用机器学习来改善木质纤维素生物质的热解，重点是机器学习算法和产品特性的预测。算法包括回归分析、人工神经网络、决策树和支持向量机。机器学习可以预测沼气、生物油和生物炭的产量、质量、表面积、反应动力学、技术经济学和生命周期评估。讨论了机器学习技术和工程应用的鲁棒性。