Abstract: Product prediction of biomass gasification is significant for accurately regulating biomass gasifiers. Machine learning (ML) based prediction methods have fast calculation speed and high fitting accuracy. However, the lack of experimental samples makes constructing a highly credible ML model challenging. To solve this problem, this paper proposes a highly effective method for predicting the distribution of biomass gasification products based on data augmentation and model migration. The method first constructs a kinetic model of biomass gasification reaction to generate simulation data and achieve data augmentation. Furthermore, it adopts a neural network to establish a pre-training model based on sufficient simulation data. Then, linear and non-linear correction networks are added based on the pre-training model, and experimental data is utilized to update the correction networks and migrate the pre-training model to the feature space adapted to the experimental data. The proposed method is used to construct a prediction model for woody biomass, and the results show that this model achieves accurate prediction for five testing samples with only four training samples, where the coefficient of determination R² is 0.98 and root mean square error is 0.64%. This proposed method demonstrates superior generalization ability and interpretability compared with the state-of-the-art methods.