What the study found
The study reports that a photovoltaic power prediction method combining improved K-means clustering, HPO-VMD, and HPO-BiLSTM was effective. The authors say the experimental and analytical results verified the method’s effectiveness and generalization.
Why the authors say this matters
The authors say photovoltaic power generation is affected by volatile meteorological factors, which can seriously influence the stability of grid connection. They present their method as a way to improve prediction accuracy under these conditions.
What the researchers tested
The researchers built a numerical prediction method for photovoltaic power generation. They used arrangement entropy combined with K-means clustering for weather typing, HPO-VMD for adaptive data decomposition, and a BiLSTM model whose parameters were adjusted using the hunter-prey algorithm (HPO).
What worked and what didn't
The abstract says the weather-typing step was used to reduce the influence of data diversity and randomness on prediction accuracy. It also says HPO-VMD was used to handle strong volatility and randomness in photovoltaic power data, and HPO was used to reduce the negative effects of poorly chosen model parameters. The abstract does not report any specific numerical results or failures.
What to keep in mind
The available summary does not give quantitative performance measures, comparison baselines, or detailed limitations. The testing described in the abstract was based on actual data from the Alice Springs site, so the stated validation is limited to that data source in the summary provided.
Key points
- The study proposes a short-term photovoltaic power prediction method combining improved K-means, HPO-VMD, and HPO-BiLSTM.
- Arrangement entropy was combined with K-means clustering to support weather typing.
- HPO-VMD was used for adaptive data decomposition to address volatility and randomness in photovoltaic power data.
- A BiLSTM model was used for numerical prediction, with HPO used to adjust model parameters.
- The abstract says experiments on Alice Springs data verified effectiveness and generalization.
Disclosure
- Research title:
- HPO-VMD-BiLSTM improved short-term photovoltaic power prediction
- Authors:
- Jianbo Li, Longhao Li, Qinjun Du, Yede Li
- Institutions:
- Shandong University of Technology, Shandong University of Technology, Shandong University of Technology, Zibo Vocational Institute
- Publication date:
- 2026-03-14
- OpenAlex record:
- View
- Image credit:
- Grendelkhan, Wikimedia Commons, CC BY-SA 4.0
Get the weekly research newsletter
Stay current with scholarly research without reading academic papers — one filtered digest, every Friday.