Scholarlink

To enhance the accuracy of photovoltaic power forecasting, this paper proposes a deep learning model, CNN-BiLSTM-Attention-KDE, optimized using the Differential Creative Search (DCS) algorithm, for short-term photovoltaic power interval prediction. The model combines a Convolutional Neural Network (CNN), Bidirectional Long Short-Term Memory Network (BiLSTM), and Attention Mechanism to extract the spatiotemporal features of photovoltaic power, capture the bidirectional dependencies within the time series, and dynamically adjust the importance of the input features. To further improve prediction accuracy, the DCS algorithm is applied to optimize the model's hyperparameters. Finally, the Kernel Density Estimation (KDE) method is employed for power interval prediction. Experimental results demonstrate that the proposed model significantly outperforms conventional methods in terms of prediction accuracy and computational efficiency, offering more reliable photovoltaic power forecasts for grid dispatch and energy management.

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

Short-Term Interval Forecasting of Photovoltaic Power Based on CNN-BiLSTM-Attention and DCS Optimization

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Chaolong Tang,
  • Jun Su,
  • Yihan Yang,
  • Zita Wang,
  • Chuzhi Huang

摘要

To enhance the accuracy of photovoltaic power forecasting, this paper proposes a deep learning model, CNN-BiLSTM-Attention-KDE, optimized using the Differential Creative Search (DCS) algorithm, for short-term photovoltaic power interval prediction. The model combines a Convolutional Neural Network (CNN), Bidirectional Long Short-Term Memory Network (BiLSTM), and Attention Mechanism to extract the spatiotemporal features of photovoltaic power, capture the bidirectional dependencies within the time series, and dynamically adjust the importance of the input features. To further improve prediction accuracy, the DCS algorithm is applied to optimize the model's hyperparameters. Finally, the Kernel Density Estimation (KDE) method is employed for power interval prediction. Experimental results demonstrate that the proposed model significantly outperforms conventional methods in terms of prediction accuracy and computational efficiency, offering more reliable photovoltaic power forecasts for grid dispatch and energy management.