A method for rapid cluster partitioning of photovoltaic plants based on prototype extraction and clustering

Abstract

The penetration rate of photovoltaic power generation keeps increasing. To address the issues of poor cluster partitioning and lengthy processing times for photovoltaic power station clusters， the paper proposes a method for rapid cluster partitioning method for photovoltaic （PV） plants based on prototype extraction and clustering. Firstly， photovoltaic data is preprocessed to eliminate differences in magnitude and dimensionality among data sets. Subsequently， influential factors on photovoltaic output power are identified using the Pearson correlation coefficient method. Random sampling， k-means++， and an improved spectral clustering method are then employed for sampling， prototype extraction， and prototype clustering of PV plants， respectively. Building upon an enumeration approach and hierarchical optimization， optimal hyperparameters for the aforementioned processes are determined. Finally， various scenarios are set up for case study comparisons， calculating both intra-cluster and inter-cluster indicators as well as clustering time metrics. Through comprehensive analysis， the effectiveness of the proposed method in addressing the rapid clustering for large-scale PV plants is validated.

Keywords

• photovoltaic power station
• improved spectral clustering algorithm
• prototype clustering
• pearson correlation coefficient