A method for evaluating surface solar radiation based on off-line reinforcement learning for non-measurable areas
By combining offline reinforcement learning and adaptive conservative Q-learning algorithms with multi-source data and uncertainty assessment, the problems of data missingness and distribution out-generalization in the assessment of surface solar radiation in unmeasured areas are solved, achieving high-precision and robust assessment results.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- INST OF ENERGY HEFEI COMPREHENSIVE NAT SCI CENT (ANHUI ENERGY LAB)
- Filing Date
- 2026-05-14
- Publication Date
- 2026-07-17
AI Technical Summary
Existing methods for assessing surface solar radiation suffer from data gaps and out-of-distribution generalization in areas without measurement, making it difficult to maintain a balance between assessment accuracy and robustness without online interaction with the real environment.
An offline reinforcement learning method is adopted. By constructing feature vectors of multi-source satellite data and ground meteorological elements, a policy network is trained to learn fusion weights. Combined with an adaptive conservative Q-learning algorithm, the conservatism is dynamically adjusted to construct a robust decision-making mechanism, and an uncertainty assessment mechanism is introduced.
High-precision assessment of surface solar radiation was achieved in areas without measurement capabilities, reducing the risk of overestimation, improving the model's adaptability and robustness, and forming a complete engineering closed-loop process applicable to photovoltaic power generation assessment and power dispatch.
Smart Images
Figure CN122416293A_ABST