Summary
Amid the construction of new power systems, traditional power supply planning faces bottlenecks like delayed demand response and over-reliance on expert experience. This study proposes the PowerPlan-GPT framework, integrating multimodal large language models, graph spatio-temporal neural networks, and adaptive evolutionary algorithms to build a fourtier intelligent generation system. Leveraging RAG technology and rule-constrained reasoning, it achieves global optimization across cost, safety, and timeliness. Piloted in
Read more Read lessJiangsu Power Grid, the framework reduced the average solution generation cycle from 48 hours to 50 minutes, raised the one-time approval rate to 55.7%, and reached 95.34% accuracy. Now deployed in 26 provinces with over 890,000 solutions generated, it provides a practical pathway for "source-grid-load-storage" coordinated optimization.
Additional informations
| Publication type | Session Materials |
|---|---|
| Reference | D2_11578_2026 |
| Publication year | |
| Publisher | CIGRE |
| Country | China, People's Republic of |
| Study committees | |
| File size | 1 MB |
| Price for non member | 30 € |
| Price for member | 30 € |
Authors
FENG Zhipeng - SGICT; CAI Xinyi - SGICT; ZHANG Pan - SGICT; ZHAO Xiqing - SGICT; CHEN Qifan - SGICT
Keywords
