Linfei Yin

The medium- and low-voltage distribution network is a critical hub connecting the transmission grid and end-users, and its power supply reliability directly determines livelihood security and socio-economic operational efficiency. Typhoon-induced strong winds and rainfall often trigger large-scale power outage risks, severely threatening power grid security and resilience. To achieve the rapid and accurate perception of outage risk areas based on building damage after typhoons, this study proposes the ResiDS-Net method, which infers distribution network outage risk levels by identifying building damage levels. An improved Dempster–Shafer evidence theory is here adopted to fuse the outputs of CM-ResNet50, Inception-V3 and DenseNet121, enhancing perception accuracy. A two-stage “coarse screening–fine judgment” framework using dual datasets is established to quickly identify large-scale suspected power outage areas from building group damage data. To address the issue that equating building damage with power outages reduces judgment accuracy, this study further develops a hierarchical building damage dataset, classifying individual buildings by damage level to achieve precise outage risk identification. Our experiments show that ResiDS-Net achieves 96.66% and 93.00% accuracy on the two datasets, 2.13% and 2.50% higher than nine comparative networks including Inception-V3. The proposed method effectively improves outage risk perception precision and provides a scientific basis for power emergency repair. It should be noted that the proposed method provides building-damage-based outage risk inference for emergency decision support, rather than the direct detection of verified actual outage status.

Accurate wind power forecasting can significantly reduce grid dispatch costs and improve system performance. Consequently, this study proposes a bimodal-driven method based on large language model and interpretable convolutional attention-gated (BD-LLICAG) network for wind power forecast. The BD-LLICAG method utilizes the dataset collected from microsensors in the Internet of Things for training and is compared with 53 classic algorithms. The BD-LLCAG has 53.38% and 32.62% less than the average absolute error metrics of a single interpretable convolutional attention-gated (ICAG) network and DeepSeek large language model, respectively, and has 26.75% less average absolute error metrics than the state-of-the-art inception -embedded attention memory fully-connected network. Moreover, through comparative analysis of ablation experiment results and the Wilcoxon signed-rank test for BD-LLICAG, the dual-model-driven wind power forecasting model—combining DeepSeek with ICAG achieved optimal prediction performance is demonstrated. Compared to the single ICAG network and DeepSeek model, the mean absolute error metrics are reduced by 53.38% and 32.62%, respectively. Finally, the combined application of Shapley values and N-BEATS in tandem is successfully applied to visualize and analyze the extent to which the input wind power feature variables influence the prediction results of the wind power predicted by BD-LLICAG.