Gradient boosting dendritic network for ultra-short-term PV power prediction

Researchers at Central South University in China have developed a new model to improve ultra-short-term photovoltaic (PV) power prediction, as detailed in their publication in Frontiers in Energy.

In an effort to enhance the efficiency of PV power generation systems, the research team led by Yuan Cao has introduced the Gradient Boosting Dendritic Network (GBDD) model. This novel ensemble prediction model is designed to address the challenges of accurate power forecasting by integrating a gradient boosting strategy with a dendritic network. The significance of this development lies in its potential to significantly reduce forecast errors by learning the relationship between forecast residuals and meteorological factors.

The background for this research stems from the increasing demand for reliable power forecasting models that can support effective intraday dispatch of PV systems. Current methodologies often fall short due to their limited use of meteorological data and challenges in model interpretation. The GBDD model overcomes these limitations by leveraging a greedy function approximation during sub-model training, allowing for the comprehensive use of meteorological data and improved model transparency.

Key findings from the study indicate that the GBDD model not only enhances PV power prediction accuracy but also offers a framework to improve the performance of other prediction models. By analyzing prediction errors and their relation to meteorological factors, the model effectively compensates for inaccuracies in existing prediction methodologies.

The research utilized a robust experimental setup, analyzing data to train the GBDD model and validate its superiority over traditional models. This approach ensures the reliability of the findings and highlights the model’s adaptability to diverse PV systems.

The implications of this research are profound for both academia and industry. The model’s ability to improve PV power prediction accuracy could lead to more efficient energy management and potentially influence future energy policies. Moreover, the GBDD model’s framework offers a pathway for developing enhanced prediction models across various sectors.

The study was supported by the National Natural Science Foundation of China and the Natural Science Foundation of Hunan Province, China.

Original source: 

https://link.springer.com/article/10.1007/s11708-024-0915-y

https://journal.hep.com.cn/fie/EN/10.1007/s11708-024-0915-y

Sharable link: https://rdcu.be/eQ1Ui