Scholarlink

<p>For effective planning and utilization of wind energy resources, accurate and precise wind speed forecasting is of the utmost importance. This study evaluates several forecasting models, including auto-regressive integrated moving average (ARIMA) and its hybrid variants: ARIMA-genetic algorithm (ARIMA-GA), ARIMA-artificial neural network (ARIMA-ANN), ARIMA-particle swarm optimization (ARIMA-PSO), ARIMA-support vector machine (ARIMA-SVM), and ARIMA-generalized autoregressive conditional heteroskedasticity (ARIMA-GARCH). In this research work, ARIMA-GA is introduced as a novel model for wind speed forecasting. The study highlights the improvement in forecast accuracy using met-mast data&#xa0;of one year from Jafarabad, Gujarat and preprocessing the data using the K-nearest neighbours (KNN) method. The hybrid model uses a genetic algorithm (GA) to improve the model&#xa0;parameters, such as auto-regressive (AR) coefficients (<i>ϕ</i>), moving average (MA) coefficients (θ<i>)</i>, and variance associated with the white noise term (σ<sup>2</sup><i>)</i>, using rigorous training and validation methods.&#xa0;The study concludes by identifying the hybrid model as providing the better forecast.&#xa0;Forecasting performance capability is evaluated by achieving the minimization of mean absolute error (MAE), mean absolute percentage error (MAPE) and root mean square error (RMSE). The&#xa0;values obtained for&#xa0;the ARIMA and ARIMA-GA models&#xa0;are 0.5148, 0.6318 and 9.338, and 0.4963, 0.6260 and 9.0679, respectively, confirm that the suggested methodology is beneficial in improving forecasting precision. For&#xa0;stakeholders involved in the wind energy industry, this verified model offers a reliable resource that will help them make well-informed decisions.</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

Advancing wind speed forecasting accuracy with a novel ARIMA-GA hybrid model and enhanced parameter optimization

  • 期刊论文

  • 发表日期：

    2026年2月20日

查看全文
  • Yasir Baig,
  • Siraj Ahmed

摘要

For effective planning and utilization of wind energy resources, accurate and precise wind speed forecasting is of the utmost importance. This study evaluates several forecasting models, including auto-regressive integrated moving average (ARIMA) and its hybrid variants: ARIMA-genetic algorithm (ARIMA-GA), ARIMA-artificial neural network (ARIMA-ANN), ARIMA-particle swarm optimization (ARIMA-PSO), ARIMA-support vector machine (ARIMA-SVM), and ARIMA-generalized autoregressive conditional heteroskedasticity (ARIMA-GARCH). In this research work, ARIMA-GA is introduced as a novel model for wind speed forecasting. The study highlights the improvement in forecast accuracy using met-mast data of one year from Jafarabad, Gujarat and preprocessing the data using the K-nearest neighbours (KNN) method. The hybrid model uses a genetic algorithm (GA) to improve the model parameters, such as auto-regressive (AR) coefficients (ϕ), moving average (MA) coefficients (θ), and variance associated with the white noise term (σ2), using rigorous training and validation methods. The study concludes by identifying the hybrid model as providing the better forecast. Forecasting performance capability is evaluated by achieving the minimization of mean absolute error (MAE), mean absolute percentage error (MAPE) and root mean square error (RMSE). The values obtained for the ARIMA and ARIMA-GA models are 0.5148, 0.6318 and 9.338, and 0.4963, 0.6260 and 9.0679, respectively, confirm that the suggested methodology is beneficial in improving forecasting precision. For stakeholders involved in the wind energy industry, this verified model offers a reliable resource that will help them make well-informed decisions.