Scholarlink

<p>Wind energy is a promising renewable resource, but its variable nature makes accurate wind speed prediction challenging. This study evaluated five Machine Learning (ML) techniques: Multiple Linear Regression, Support Vector Machines, Random Forests, Stochastic Gradient Boosting (SGBoost), and Artificial Neural Networks (ANNs) to forecast hourly wind speeds at two locations in Hyderabad, India (HL1—17.4° N, 78.5° E, elevation ~ 495&#xa0;m; HL2—17.4° N, 78.4° E, elevation ~ 542&#xa0;m). In parallel, a long-term wind resource assessment (2014–2023) showed that HL1 and HL2 exhibit mean wind speeds of 5.10 and 5.06&#xa0;m/s, with Weibull scale factors of ~ 5.75 and ~ 5.71&#xa0;m/s and corresponding wind power densities of 137.4 and 131.8&#xa0;W/m², respectively, reflecting promising wind conditions for regional renewable-energy development. To enhance predictive accuracy, key meteorological parameters, such as surface air temperature, mean sea level pressure, soil temperature, long-wave radiation, surface pressure, skin temperature, cloud cover, and soil moisture, were incorporated. Model performance was assessed using Root Mean Square Error (RMSE), Correlation Coefficient (CC), Mean Absolute Error, Index of Agreement (IOA), BIAS, and Standard Deviation. Results indicated that all models provided statistically reliable predictions; however, ANN consistently outperformed the others (RMSE = 0.7, CC = 0.96, IOA = 0.98), with bias tightly constrained between − 1.5 and + 1.5&#xa0;m/s, compared to the wider variability of alternative approaches. Seasonal analysis further confirmed ANN’s robustness across all periods, while SGBoost delivered complementary performance during the monsoon (JJAS) season. Overall, these findings demonstrated the potential of ML techniques, particularly ANNs, for accurate wind speed forecasting, providing critical insights for renewable energy planning and regional wind power 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

Short-term wind speed forecasting using machine learning models with integrated long-term wind resource assessment

  • 期刊论文

  • 发表日期：

    2026年2月18日

查看全文
  • Hampika Gorla,
  • N. Venkatram,
  • G. Ch. Satyanarayana,
  • Sambasivarao Velivelli,
  • L. V. Narsimha Prasad

摘要

Wind energy is a promising renewable resource, but its variable nature makes accurate wind speed prediction challenging. This study evaluated five Machine Learning (ML) techniques: Multiple Linear Regression, Support Vector Machines, Random Forests, Stochastic Gradient Boosting (SGBoost), and Artificial Neural Networks (ANNs) to forecast hourly wind speeds at two locations in Hyderabad, India (HL1—17.4° N, 78.5° E, elevation ~ 495 m; HL2—17.4° N, 78.4° E, elevation ~ 542 m). In parallel, a long-term wind resource assessment (2014–2023) showed that HL1 and HL2 exhibit mean wind speeds of 5.10 and 5.06 m/s, with Weibull scale factors of ~ 5.75 and ~ 5.71 m/s and corresponding wind power densities of 137.4 and 131.8 W/m², respectively, reflecting promising wind conditions for regional renewable-energy development. To enhance predictive accuracy, key meteorological parameters, such as surface air temperature, mean sea level pressure, soil temperature, long-wave radiation, surface pressure, skin temperature, cloud cover, and soil moisture, were incorporated. Model performance was assessed using Root Mean Square Error (RMSE), Correlation Coefficient (CC), Mean Absolute Error, Index of Agreement (IOA), BIAS, and Standard Deviation. Results indicated that all models provided statistically reliable predictions; however, ANN consistently outperformed the others (RMSE = 0.7, CC = 0.96, IOA = 0.98), with bias tightly constrained between − 1.5 and + 1.5 m/s, compared to the wider variability of alternative approaches. Seasonal analysis further confirmed ANN’s robustness across all periods, while SGBoost delivered complementary performance during the monsoon (JJAS) season. Overall, these findings demonstrated the potential of ML techniques, particularly ANNs, for accurate wind speed forecasting, providing critical insights for renewable energy planning and regional wind power development.