A Distributed Photovoltaic Cluster Partitioning Method Based on Spatiotemporal Scalability and Feature Correlation Minimization (SSFCM)

We proposed a distributed photovoltaic (PV) cluster partitioning method based on spatiotemporal scalability and feature correlation minimization. By considering the temporal and spatial characteristics of distributed PV output curves, we introduced the capacity factor and dynamic time windows into the Dynamic Time Warping (DTW) calculation process, decomposing the output curves into multiple time windows. Building on the K-Means algorithm, we proposed a feature correlation minimization principle to guide the initialization of cluster centers, grouping distributed PV systems with similar output characteristics into the same cluster. Finally, by integrating meteorological data and historical PV output data, we constructed a MQ-RNN prediction model to effectively forecast PV output. The case study results demonstrated that the proposed method in this paper is capable of handling scenarios with inconsistent data, such as long-term intermittent missing data. It can reduce the clustering error of distributed photovoltaic systems, thereby improving the accuracy of power prediction for distributed photovoltaics.