(1) Objective: To improve forest fire prevention, this study provides a reference for forest fire risk assessment in Sichuan Province. (2) Methods: This research focuses on various forest vegetation types in Sichuan Province. Given data from 6848 sample plots, five machine learning models—random forest, extreme gradient boosting (XGBoost), k-nearest neighbors, support vector machine, and stacking ensemble (Stacking)—were employed. Bayesian optimization was utilized for hyperparameter tuning, resulting in machine learning models for predicting forest fuel loads (FLs) across five different vegetation types. (3) Results: The FL model incorporates not only vegetation characteristics but also site conditions and climate data. Feature importance analysis indicated that structural factors (e.g., canopy closure, diameter at breast height, and tree height) dominated in cold broadleaf, subtropical broadleaf, and subtropical mixed forests, while climate factors (e.g., mean annual temperature and temperature seasonality) were more influential in cold coniferous and subtropical coniferous forests. Machine learning-based FL models outperform the multiple stepwise regression model in both fitting ability and prediction accuracy. The XGBoost model performed best for cold coniferous, cold broadleaf, subtropical broadleaf, and subtropical mixed forests, with coefficient of determination (R²) values of 0.79, 0.85, 0.81, and 0.83, respectively. The Stacking model excelled in subtropical coniferous forests, achieving an R² value of 0.82. (4) Conclusions: This study establishes a theoretical foundation for predicting forest fuel capacity in Sichuan Province. It is recommended that the XGBoost model be applied to predict fuel loads (FLs) in cold coniferous forests, cold broadleaf forests, subtropical broadleaf forests, and subtropical mixed forests, while the Stacking model is suggested for predicting FLs in subtropical coniferous forests. Furthermore, this research offers theoretical support for forest fuel management, forest fire risk assessment, and forest fire prevention and control in Sichuan Province.

Arc fault detection is critical for ensuring the safety and reliability of electrical power systems, as it helps prevent fires and equipment damage. Data models demonstrate significant advantages across various fields and have become a focal point of arc fault detection. However, the generalization ability under small sample conditions and interpretability of these models remain key challenges in this field. This article presents an end-to-end arc fault detection method that integrates experience. First, based on current waveforms, three types of artificial experience are summarized. Then, a one-dimensional adaptive neural decision tree (1D-ANT) based arc fault detection method is explored by combining neural networks and decision tree to perform feature extracting, path selecting, and label classifying. Finally, by embedding experience into the 1D-ANT to guide model training, the trained model is capable of effectively detecting arc faults across various load types. The results show that the proposed method achieves a high fault detection performance on datasets of different sizes, and is compared with other reported methods to verify its superiority. Moreover, it can visualize the decision-making process.

To address the challenges of vegetation interference, background confusion, and road fragmentation caused by the narrow and elongated structures of forest roads in complex remote sensing imagery, this study proposed an improved U-Net-based model, namely HAA-UNet, for automatic forest road extraction. The proposed model integrates a VGG16 encoder, an Atrous Spatial Pyramid Pooling (ASPP) module, and a Hybrid Dilated Convolution (HDCC) module to enhance feature representation and spatial detail reconstruction of road targets under complex forest environments. Experimental results demonstrated that HAA-UNet achieved superior segmentation performance on the forest road dataset of Xichang City, Sichuan Province, with Precision, Recall, F1-score, and mIoU values of 87.24%, 87.83%, 87.53%, and 79.85%, respectively, outperforming all comparison models. These results indicate that the proposed method effectively improves road continuity and boundary delineation in complex forest scenes. Furthermore, the extracted road information was integrated into the Forest Fire Risk Index (FFRI) assessment framework, demonstrating that accurate road data can improve the spatial characterization of fire risk and provide reliable data support for forest fire risk assessment and forest resource management.