Scholarlink

<p>The proposed control strategy aims to get the most power possible from a variety of energy sources in an isolated AC Microgrid by keeping a steady energy surplus without needing extra loads or special communication infrastructure. It uses the Microgrid’s electrical frequency, which is usually kept close to 50&#xa0;Hz, as a built-in way for distributed energy resources (DERs) and power converters to work together. A Long Short-Term Memory (LSTM) neural network is used to predict short-term renewable generation and load demand in order to make the system more responsive. These predictions let you make changes to the output of each source ahead of time, which makes frequency-based coordination more accurate and stable. The results of the simulation show that the method can use achieving 92% renewable energy penetration level during the simulation period, compared to 76% of the time with traditional droop-based control. Frequency changes are limited to ± 0.15&#xa0;Hz, which keeps performance stable even when the load changes. The method also cuts battery cycling by 28%, which makes energy storage systems last longer. The strategy makes it easy to combine solar, wind, and storage units because it responds quickly, in less than 150 milliseconds. Also, operational costs go down by about 18%, and power is always available. Adding LSTM-based forecasting makes the system much more reliable and flexible. This makes the proposed method a strong, scalable, and long-lasting solution for future smart Microgrids, especially in remote and off-grid settings.</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

AI-enhanced operation and control of isolated AC micro-grids with renewable energy and energy storage systems

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2025年12月24日

查看全文
  • Sathishkumar Ramasamy,
  • Sriram Anbalagan,
  • D. Manikandan,
  • Krishnaraj Ramaswamy

摘要

The proposed control strategy aims to get the most power possible from a variety of energy sources in an isolated AC Microgrid by keeping a steady energy surplus without needing extra loads or special communication infrastructure. It uses the Microgrid’s electrical frequency, which is usually kept close to 50 Hz, as a built-in way for distributed energy resources (DERs) and power converters to work together. A Long Short-Term Memory (LSTM) neural network is used to predict short-term renewable generation and load demand in order to make the system more responsive. These predictions let you make changes to the output of each source ahead of time, which makes frequency-based coordination more accurate and stable. The results of the simulation show that the method can use achieving 92% renewable energy penetration level during the simulation period, compared to 76% of the time with traditional droop-based control. Frequency changes are limited to ± 0.15 Hz, which keeps performance stable even when the load changes. The method also cuts battery cycling by 28%, which makes energy storage systems last longer. The strategy makes it easy to combine solar, wind, and storage units because it responds quickly, in less than 150 milliseconds. Also, operational costs go down by about 18%, and power is always available. Adding LSTM-based forecasting makes the system much more reliable and flexible. This makes the proposed method a strong, scalable, and long-lasting solution for future smart Microgrids, especially in remote and off-grid settings.