Scholarlink

Due to the unique geographical location of rural areas, distributed PV systems are more dispersed. This leads to difficulties in mining the spatio-temporal characteristics of their output. To overcome this challenge, a graph-temporal convolutional neural network prediction method based on spatio-temporal correlation is proposed. The distributed PV system is considered a class I cell nodes. Its historical output data serves as the feature information of the node, while the topology of its spatial location is used as an edge to construct a spatio-temporal information graph of the system’s output. The adjacency matrix is determined using the spatial correlation of distributed PV systems. The spatio-temporal information graph and the spatial output is input into the graph-temporal convolutional network for training and prediction. This paper uses PV power generation data in a rural area as an example to compare with existing methods and verify the effectiveness of the proposed method.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Rural Distributed Photovoltaic Spatial Generation Forecasting Based on Graph Convolutional Neural Networks with Temporal-Spatial Correlation

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Milu Zhou,
  • Shuai Wang,
  • Xiaoyan Yang

摘要

Due to the unique geographical location of rural areas, distributed PV systems are more dispersed. This leads to difficulties in mining the spatio-temporal characteristics of their output. To overcome this challenge, a graph-temporal convolutional neural network prediction method based on spatio-temporal correlation is proposed. The distributed PV system is considered a class I cell nodes. Its historical output data serves as the feature information of the node, while the topology of its spatial location is used as an edge to construct a spatio-temporal information graph of the system’s output. The adjacency matrix is determined using the spatial correlation of distributed PV systems. The spatio-temporal information graph and the spatial output is input into the graph-temporal convolutional network for training and prediction. This paper uses PV power generation data in a rural area as an example to compare with existing methods and verify the effectiveness of the proposed method.