Wide-area photovoltaic power prediction method and system based on spatiotemporal residual feature fusion

By constructing a two-level cascaded prediction framework based on the fusion of spatiotemporal residual features, the problem of insufficient accuracy in photovoltaic power prediction over a wide area is solved, and high-precision prediction of regional photovoltaic output is achieved, especially with adaptive improvement under sudden weather conditions.

CN122026337BActive Publication Date: 2026-06-26SHANDONG JIANZHU UNIV
View PDF 2 Cites 0 Cited by

Patent Information

Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
SHANDONG JIANZHU UNIV
Filing Date
2026-04-14
Publication Date
2026-06-26

AI Technical Summary

Technical Problem

Existing photovoltaic power prediction methods fail to fully exploit the regional spatiotemporal patterns contained in the residuals over a wide area, lack modeling for geographically dispersed smoothing effects, and are not closely integrated with the physical processes of macro-meteorological driving regional total power output, resulting in insufficient prediction accuracy.

Method used

A two-level cascaded prediction framework based on spatiotemporal residual feature fusion is constructed. By acquiring multi-source heterogeneous data, regional macro-meteorological features are extracted, and multiple first-level prediction models are used for parallel prediction and then aggregated to construct spatiotemporal residual features. These features are then refined by combining a second-level correction model to finally generate the predicted value of the total photovoltaic power in the region.

Benefits of technology

It achieves in-depth modeling of regional photovoltaic power output patterns, improving prediction accuracy. In particular, it has self-learning and adaptive capabilities under sudden weather conditions, dynamically improving prediction accuracy.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN122026337B_ABST
    Figure CN122026337B_ABST
Patent Text Reader

Abstract

The application belongs to the field of photovoltaic prediction, and provides a wide-area photovoltaic power prediction method and system based on spatiotemporal residual feature fusion, which comprises the following steps: based on regional macro weather characteristics, a plurality of trained first-level prediction models are used to perform prediction in parallel, and then the first-level aggregated prediction value and a prediction residual sequence are obtained; according to the prediction residual sequence, corresponding spatiotemporal residual features are constructed, the spatiotemporal residual features are fused with the time-aligned regional macro weather characteristics to obtain an enhanced feature set; based on the enhanced feature set, a trained second-level correction model is used to perform prediction to obtain a final predicted power correction amount; and the sum of the final predicted power correction amount and the first-level aggregated prediction value is taken as the regional photovoltaic total power prediction value. The application explicitly learns and corrects systematic prediction bias caused by regional geographical dispersion, thereby reducing dependence on underlying massive data.
Need to check novelty before this filing date? Find Prior Art