Scholarlink

<p>In this paper, a photovoltaic power (PV) forecasting model based on a tensor calculation method, called tensor echo state network (T-ESN), is proposed to solve the influence of invalid data in the original data of photovoltaic power on the forecasting accuracy. First, the state dimension of each reservoir is expanded using the idea of tensor expansion to better learn the hidden features in the data; second, the expanded states are weighted and contracted by the principle of tensor contraction, and the weights are assigned to the output states of multiple reservoirs to obtain the final network state. To illustrate the reliability of the improved network, the stability proof of T-ESN is given, and the adverse effect of parameters on the model performance is eliminated by the white shark optimization (WSO) algorithm. Finally, to verify the forecasting effect of the proposed artificial intelligence method, three specific example simulations are carried out, and the results show that the performance improvement of T-ESN compared to the traditional model can be up to more than 90%. This indicates that the PV power generation forecasting model based on T-ESN has significant advantages in terms of practicality, generalization, and accuracy.</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 photovoltaic power forecasting model based on tensor echo state network

  • 期刊论文

  • 发表日期：

    2026年7月9日

查看全文
  • Xianshuang Yao,
  • Hailong Xu,
  • Jihan Sun,
  • Xiaoli Li,
  • Qingchuan Ma

摘要

In this paper, a photovoltaic power (PV) forecasting model based on a tensor calculation method, called tensor echo state network (T-ESN), is proposed to solve the influence of invalid data in the original data of photovoltaic power on the forecasting accuracy. First, the state dimension of each reservoir is expanded using the idea of tensor expansion to better learn the hidden features in the data; second, the expanded states are weighted and contracted by the principle of tensor contraction, and the weights are assigned to the output states of multiple reservoirs to obtain the final network state. To illustrate the reliability of the improved network, the stability proof of T-ESN is given, and the adverse effect of parameters on the model performance is eliminated by the white shark optimization (WSO) algorithm. Finally, to verify the forecasting effect of the proposed artificial intelligence method, three specific example simulations are carried out, and the results show that the performance improvement of T-ESN compared to the traditional model can be up to more than 90%. This indicates that the PV power generation forecasting model based on T-ESN has significant advantages in terms of practicality, generalization, and accuracy.