N. Lu, X. Fu, P. Zhang, D. Qiu, H. Badihi, M. Abdel-Salam, and H. Gu, “Enhancing representative photovoltaic scenario extraction for multiple power stations with a shared-weight and adaptively fused graph clustering method,” Applied Energy, vol. 406, 2026.
New graph clustering method for representative photovoltaic scenario extraction published in Applied Energy
A new research article published in Applied Energy proposes a novel graph clustering framework for extracting representative photovoltaic (PV) power generation scenarios from multiple power stations.
The method introduces a shared-weight and adaptively fused graph clustering model that effectively captures correlations among distributed PV stations while preserving output variability. By incorporating a rank-constrained Laplacian matrix and an efficient alternating optimization algorithm, the framework can automatically determine the optimal number of clusters with fast convergence.
Extensive theoretical analysis and simulation results demonstrate strong robustness to noise and data irregularities, as well as low sensitivity to model parameters. Validation using real PV data and probabilistic power flow analysis further highlights the potential of the proposed method for power system planning under high renewable energy penetration.