Scholarlink

Accurate prediction of photovoltaic (PV) power contributes to rational scheduling of photovoltaic grid-connection and reliable grid operation. In response to complexity of PV power, this study proposes a PV power prediction model based on similar day clustering and a Secretary Bird Optimisation Algorithm (SBOA) optimised CNN-BiLSTM-Attention framework. Firstly, the main meteorological features were selected using the Pearson coefficient (PCC); then, use the K-means++ algorithm to cluster the historical data into three weather types; and utilize the meteorological data from the forecasted day, along with meteorological and power data from similar historical days, as training samples to train the model; optimizing the initial learning rate, L2 regularization parameters, and number of hidden neurons of the composite model using SBOA. Finally, taking the dataset of a PV power plant in Xinjiang as an example, and simulation results demonstrate that the proposed combined model can predict photovoltaic power generation under various weather conditions with greater accuracy compared to other predictive models.

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 Photovoltaic Power Prediction Based on Similar Day Clustering and Combined SBOA-CNN-BiLSTM-Attention Model

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Youwei Li,
  • Fang Wang,
  • Weiguang Gu

摘要

Accurate prediction of photovoltaic (PV) power contributes to rational scheduling of photovoltaic grid-connection and reliable grid operation. In response to complexity of PV power, this study proposes a PV power prediction model based on similar day clustering and a Secretary Bird Optimisation Algorithm (SBOA) optimised CNN-BiLSTM-Attention framework. Firstly, the main meteorological features were selected using the Pearson coefficient (PCC); then, use the K-means++ algorithm to cluster the historical data into three weather types; and utilize the meteorological data from the forecasted day, along with meteorological and power data from similar historical days, as training samples to train the model; optimizing the initial learning rate, L2 regularization parameters, and number of hidden neurons of the composite model using SBOA. Finally, taking the dataset of a PV power plant in Xinjiang as an example, and simulation results demonstrate that the proposed combined model can predict photovoltaic power generation under various weather conditions with greater accuracy compared to other predictive models.