Scholarlink

<p>The photovoltaic (PV) power generation process is susceptible to the interference of meteorological factors with volatility, which seriously impacts the stability of PV grid connection. To solve the above problems, this paper proposes a numerical prediction method for PV power generation with improved K-means, HPO-VMD and HPO-BiLSTM. Firstly, to reduce the influence of data diversity and randomness on the PV power prediction accuracy, the arrangement entropy is combined with K-means clustering algorithm to achieve weather typing under different meteorological factors. Then, for the existence of strong volatility and randomness of PV power data, HPO-VMD data decomposition method is proposed to achieve adaptive data decomposition and improve the smoothness of PV power values. Subsequently, the PV power numerical prediction model is constructed based on the BiLSTM method. Meanwhile, to reduce the adverse effects of improper selection of model parameters on the model prediction accuracy, the hunter–prey algorithm (HPO) is used to achieve the model parameter rectification. Finally, simulation experiments of the proposed method are conducted based on actual data from Alice Springs site, and the experimental and analytical results verify the effectiveness and generalization of the proposed method.</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

Research on short-term prediction method of photovoltaic power based on HPO-VMD-BiLSTM

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年3月14日

查看全文
  • Jianbo Li,
  • Longhao Li,
  • Qinjun Du,
  • Yede Li

摘要

The photovoltaic (PV) power generation process is susceptible to the interference of meteorological factors with volatility, which seriously impacts the stability of PV grid connection. To solve the above problems, this paper proposes a numerical prediction method for PV power generation with improved K-means, HPO-VMD and HPO-BiLSTM. Firstly, to reduce the influence of data diversity and randomness on the PV power prediction accuracy, the arrangement entropy is combined with K-means clustering algorithm to achieve weather typing under different meteorological factors. Then, for the existence of strong volatility and randomness of PV power data, HPO-VMD data decomposition method is proposed to achieve adaptive data decomposition and improve the smoothness of PV power values. Subsequently, the PV power numerical prediction model is constructed based on the BiLSTM method. Meanwhile, to reduce the adverse effects of improper selection of model parameters on the model prediction accuracy, the hunter–prey algorithm (HPO) is used to achieve the model parameter rectification. Finally, simulation experiments of the proposed method are conducted based on actual data from Alice Springs site, and the experimental and analytical results verify the effectiveness and generalization of the proposed method.