Scholarlink

Addressing climate change, cutting greenhouse gas emissions, and achieving global sustainability objectives all depend on the switch to renewable energy. Solar energy is unique among renewable energy sources because of its extensive availability and substantial potential for both environmental and economic advantages. However, there are special difficulties in incorporating solar energy into traditional grid infrastructures, especially when it comes to handling and forecasting time-series data that is inherently variable. Aiming to successfully tackle these issues, this work presents a unique strategy that combines gated recurrent units (GRUs) with a modified sinh-cosh optimization method (SCHO). Hyperparameter optimization, which is essential for maximizing model performance and guaranteeing reliable predictions, is a major area of attention for this work. The suggested approach effectively traverses intricate hyperparameter spaces by utilizing the modified SCHO, improving model stability and accuracy. Simulations using real-world data show encouraging results; the top models achieve a mean absolute percentage error (MAPE) as low as 9.1%, indicating the efficacy of the method.

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

Photovoltaic Power Plant Production Prediction via Modified Metaheuristic Optimized Gated Recurrent Networks

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Marko Mihajlovic,
  • Snezana Malisic,
  • Vico Zeljkovic,
  • Luka Jovanovic,
  • Stefan Ivanovic,
  • Antonijevic Milos,
  • Miodrag Zivkovic,
  • Nebojsa Bacanin

摘要

Addressing climate change, cutting greenhouse gas emissions, and achieving global sustainability objectives all depend on the switch to renewable energy. Solar energy is unique among renewable energy sources because of its extensive availability and substantial potential for both environmental and economic advantages. However, there are special difficulties in incorporating solar energy into traditional grid infrastructures, especially when it comes to handling and forecasting time-series data that is inherently variable. Aiming to successfully tackle these issues, this work presents a unique strategy that combines gated recurrent units (GRUs) with a modified sinh-cosh optimization method (SCHO). Hyperparameter optimization, which is essential for maximizing model performance and guaranteeing reliable predictions, is a major area of attention for this work. The suggested approach effectively traverses intricate hyperparameter spaces by utilizing the modified SCHO, improving model stability and accuracy. Simulations using real-world data show encouraging results; the top models achieve a mean absolute percentage error (MAPE) as low as 9.1%, indicating the efficacy of the method.