Scholarlink

<p>Photovoltaic energy development has effectively mitigated energy crises and accelerated global carbon neutrality efforts. However, the increasing photovoltaic installed capacity poses significant challenges to grid scheduling systems. Photovoltaic power forecasting techniques provide crucial basis to formulate scheduling plans, thereby alleviating scheduling pressures. Yet, existing photovoltaic power prediction algorithms have shown unstable performance in complex weather conditions. Therefore, this paper proposes a short-term photovoltaic power prediction method based on the nearest clear sky day decomposition and temporal convolutional network (TCN). The method identifies the photovoltaic output on the nearest clear sky day to the target day and decomposes the photovoltaic power waveform based on the clear sky component removal. TCN combines the feature extraction capabilities of convolutional neural networks (CNNs) with the temporal information mining abilities of sequence-based neural networks like recurrent neural networks (RNNs) and long short-term memory networks (LSTMs), making it suitable for capturing the relationships between various meteorological features and photovoltaic output. Using the Alice Springs dataset in Australia as a case study, the algorithm conducts experiments under different seasons and weather conditions, comparing its performance against other models using metrics such as mean absolute error (MAE), root mean square error (RMSE), and coefficient of determination (R2).</p>

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

A Short-term photovoltaic power prediction method based on the nearest clear sky day decomposition and temporal convolutional network

  • 期刊论文

  • 发表日期：

    2025年4月18日

查看全文
  • Runxin Zhang,
  • Chengxin Pang,
  • Xiaoguang Zhu,
  • Feng Gao,
  • Pengyi Jiang

摘要

Photovoltaic energy development has effectively mitigated energy crises and accelerated global carbon neutrality efforts. However, the increasing photovoltaic installed capacity poses significant challenges to grid scheduling systems. Photovoltaic power forecasting techniques provide crucial basis to formulate scheduling plans, thereby alleviating scheduling pressures. Yet, existing photovoltaic power prediction algorithms have shown unstable performance in complex weather conditions. Therefore, this paper proposes a short-term photovoltaic power prediction method based on the nearest clear sky day decomposition and temporal convolutional network (TCN). The method identifies the photovoltaic output on the nearest clear sky day to the target day and decomposes the photovoltaic power waveform based on the clear sky component removal. TCN combines the feature extraction capabilities of convolutional neural networks (CNNs) with the temporal information mining abilities of sequence-based neural networks like recurrent neural networks (RNNs) and long short-term memory networks (LSTMs), making it suitable for capturing the relationships between various meteorological features and photovoltaic output. Using the Alice Springs dataset in Australia as a case study, the algorithm conducts experiments under different seasons and weather conditions, comparing its performance against other models using metrics such as mean absolute error (MAE), root mean square error (RMSE), and coefficient of determination (R2).