Open Access

ARTICLE

A Hybrid Forecasting Method for Wind and Photovoltaic Power Generation Considering Shared Information

Jinchuan Wang, Mingzhe Li, Xiaohong Hao^*, Tianchi Wang
School of Energy and Power Engineering, University of Shanghai for Science and Technology, Shanghai, China
* Corresponding Author: Xiaohong Hao. Email:
(This article belongs to the Special Issue: AI and Advanced Computational Techniques for Sustainable Renewable Energy Systems)

Energy Engineering https://doi.org/10.32604/ee.2026.081875

Received 10 March 2026; Accepted 27 April 2026; Published online 18 May 2026

Abstract

The strong stochasticity and spatiotemporal coupling of wind and photovoltaic (PV) power generation necessitate joint forecasting to improve the scheduling accuracy of modern power systems. However, existing approaches fail to accurately quantify the shared information between wind and PV outputs, limiting prediction performance. To address this challenge, this paper proposes a mutual information–driven dual-path adaptive collaborative forecasting framework. First, sliding-window mutual information is used to assess the local correlation between wind and PV time series, enabling partitioning of historical data into high- and low-correlation subsets. For the high-correlation subset, a multi-task bidirectional long short-term memory (BiLSTM) model is developed to facilitate joint feature learning. For the low-correlation subset, dynamic time warping (DTW) distance is employed to measure sequence similarity, followed by hierarchical clustering to identify homogeneous scenario clusters; a dedicated BiLSTM model is then trained for each cluster. During forecasting, the mutual information value of the input sequence dynamically selects the appropriate prediction path, and a nearest-neighbor routing mechanism invokes the optimal sub-model to generate the final forecast. By integrating scenario-adaptive data segmentation with group-specific modeling, the proposed method offers a novel and effective approach to collaborative forecasting of wind and solar power.

---

Keywords

New energy; wind-PV forecasting; hybrid model

---