Scholarlink

<p>Energy consumption is crucial to Vietnam’s growth. With rapid expansion, accurate forecasting is key for effective policy and sustainable development. This study proposes a hybrid model–ANN-GP-GM (1,1)–to improve the traditional Grey Model GM (1,1), which suffers from low precision due to reliance on background values. ANN(short for Artificial Neural Network), is a model designed to learn and recognize data patterns. In this study, it captures complex patterns and adapts to data changes. GP(short for Genetic Programming), is an evolutionary algorithm that mimics natural selection to optimize programs. Specifically, GP predicts residual signs, which is aimed to refine the GM (1,1) parameters and enhance forecasting accuracy. By integrating ANN to learn complex patterns and GP to optimize parameters, the new model enhances forecasting accuracy. The study compares the ANN-GP-GM (1,1) model with several forecasting approaches, including GM (1,1), MLP-GM (1,1) (MLP short for Multi-Layer Perceptron), GP-GM (1,1), NN-GM (1,1) (NN, short for Neural Network), NN-MLP-GM (1,1) and NN-GP-GM (1,1) for analyzing three real-world annual energy demand cases, using mean absolute percentage error (MAPE) as the evaluation metric. Results show that ANN-GP-GM (1,1) achieves the lowest MAPE in the first two cases (3.27% and 1.10%) and the second lowest (1.28%) in the third, with only a small gap from the best (0.53%). This demonstrates its strong and consistent performance. Compared to other models, it proves more accurate for forecasting energy demands. By providing better forecasts, it aids the government in planning investments, especially in renewable energy development.</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

An improved residual grey model integrated with artificial neural networks and genetic programming for predicting vietnam energy consumption

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年2月7日

查看全文
  • Shih-Hsien Tseng,
  • Yi-An Ko,
  • Cai-Jia Wu,
  • Le Thi Nhat Nguyen

摘要

Energy consumption is crucial to Vietnam’s growth. With rapid expansion, accurate forecasting is key for effective policy and sustainable development. This study proposes a hybrid model–ANN-GP-GM (1,1)–to improve the traditional Grey Model GM (1,1), which suffers from low precision due to reliance on background values. ANN(short for Artificial Neural Network), is a model designed to learn and recognize data patterns. In this study, it captures complex patterns and adapts to data changes. GP(short for Genetic Programming), is an evolutionary algorithm that mimics natural selection to optimize programs. Specifically, GP predicts residual signs, which is aimed to refine the GM (1,1) parameters and enhance forecasting accuracy. By integrating ANN to learn complex patterns and GP to optimize parameters, the new model enhances forecasting accuracy. The study compares the ANN-GP-GM (1,1) model with several forecasting approaches, including GM (1,1), MLP-GM (1,1) (MLP short for Multi-Layer Perceptron), GP-GM (1,1), NN-GM (1,1) (NN, short for Neural Network), NN-MLP-GM (1,1) and NN-GP-GM (1,1) for analyzing three real-world annual energy demand cases, using mean absolute percentage error (MAPE) as the evaluation metric. Results show that ANN-GP-GM (1,1) achieves the lowest MAPE in the first two cases (3.27% and 1.10%) and the second lowest (1.28%) in the third, with only a small gap from the best (0.53%). This demonstrates its strong and consistent performance. Compared to other models, it proves more accurate for forecasting energy demands. By providing better forecasts, it aids the government in planning investments, especially in renewable energy development.