Research Hilights
Generative AI for ESG Reporting (生成式 AI 永續報告)
建立可擴充的 ESG/RAG 工作流程，用於 GRI/永續報告內容生成與一致性檢核。
We develop an extensible RAG workflow for ESG reporting generation and consistency checking.
- RAG：文件分段、向量索引、引用對齊（traceability）
- 自動化：章節架構、摘要、重點表格與引用草稿
- 應用：企業 ESG 報告、政策分析、教學案例
Renewable Energy Forecasting (再生能源預測)
針對太陽光電/再生能源功率預測，整合多任務學習與不確定度分析。
Multi-task learning with uncertainty-aware forecasting for renewable generation.
- MTL / DL：LSTM、Transformer、Attention
- 評估：MAPE/MAE/RMSE、區間預測與可靠度
- 應用：智慧監控、T-REC 驗證機制改善
The Active BBU: Dynamic Power Orchestration for Stable and Efficient ORv3 AI Racks
This research has garnered significant attention in the OCP community, providing an innovative solution for AI data center power management, effectively addressing the dual challenges of idle computing resources and grid stability.
- Solving stranded AI rack capacity issues
- Meeting grid PFAPR compliance requirements
- Innovative Active BBU technology
- Dynamic power orchestration algorithms