Saved in:
| Main Author: | |
|---|---|
| Format: | Preprint |
| Published: |
2025
|
| Subjects: | |
| Online Access: | https://arxiv.org/abs/2501.08531 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
Table of Contents:
- This paper presents interval prediction methodology to address limitations in existing evaluation indicators and improve prediction accuracy and reliability. First, new evaluation indicators are proposed to comprehensively assess interval prediction methods, considering both all-sample and single-sample scenarios. Second, a novel Pattern-Diversity Conditional Time-Series Generative Adversarial Network (PDCTSGAN) is introduced to generate realistic scenarios, enabling a new interval prediction approach based on scenario generation. The PDCTSGAN model innovatively incorporates modifications to random noise inputs, allowing the generation of pattern-diverse realistic scenarios. These scenarios are further utilized to construct multiple interval patterns with high coverage probability and low average width. The effectiveness of the proposed methodology is demonstrated through comprehensive case studies. The paper concludes by highlighting future research directions to further enhance interval prediction methods.