M. H. ElTaweel, S. C. Alfaro, G. Siour, A. Coman, S. M. Robaa, M. M. Abdel Wahab
{"title":"基于ML树算法的地面风预测与预报","authors":"M. H. ElTaweel, S. C. Alfaro, G. Siour, A. Coman, S. M. Robaa, M. M. Abdel Wahab","doi":"10.1007/s00703-023-00999-6","DOIUrl":null,"url":null,"abstract":"<p>This study focuses on the importance of reliable surface wind forecasts for various sectors, particularly energy production. Traditional numerical weather prediction models are facing limitations and increasing complexity, leading to the development of machine learning models as alternatives or supplements. The research consists of two stages. In the first stage, the ERA5 database is used to evaluate the long-term performance of different combinations of features and two tree-based algorithms for predicting surface wind characteristics (speed and direction) in Cairo. The XGBoost algorithm slightly outperforms the Random Forest algorithm, especially when combined with appropriate feature selection. Even three years after the training period, the results remain very good, with an <i>RMSE</i> of 0.59 m/s, <i>rRMSE</i> of 17%, and <i>R</i><sup><i>2</i></sup> of 0.84. The second stage assesses the multivariate approach's ability to forecast wind speed evolution at different time horizons (1–12 h) during a week characterized by significant wind dynamics. The forecasts demonstrate excellent agreement with observations at a 1-h time horizon, with an <i>RMSE</i> of 0.35 m/s, <i>rRMSE</i> of 7.6%, and <i>R</i><sup><i>2</i></sup> of 0.98, surpassing or comparable to other literature results. However, as the time lag increases, the <i>RMSE</i> (0.86, 1.14, and 1.51 m/s for 3, 6, and 12 h, respectively) and <i>rRMSE</i> (18.7%, 24.8%, and 32.9% for 3, 6, and 12 h, respectively) also increase, while <i>R</i><sup><i>2</i></sup> decreases (0.86, 0.79, and 0.60). Furthermore, the wind variations' amplitude is underestimated. To address this bias, a simple correction method is proposed.</p>","PeriodicalId":51132,"journal":{"name":"Meteorology and Atmospheric Physics","volume":null,"pages":null},"PeriodicalIF":1.9000,"publicationDate":"2023-11-20","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"0","resultStr":"{\"title\":\"Prediction and forecast of surface wind using ML tree-based algorithms\",\"authors\":\"M. H. ElTaweel, S. C. Alfaro, G. Siour, A. Coman, S. M. Robaa, M. M. Abdel Wahab\",\"doi\":\"10.1007/s00703-023-00999-6\",\"DOIUrl\":null,\"url\":null,\"abstract\":\"<p>This study focuses on the importance of reliable surface wind forecasts for various sectors, particularly energy production. Traditional numerical weather prediction models are facing limitations and increasing complexity, leading to the development of machine learning models as alternatives or supplements. The research consists of two stages. In the first stage, the ERA5 database is used to evaluate the long-term performance of different combinations of features and two tree-based algorithms for predicting surface wind characteristics (speed and direction) in Cairo. The XGBoost algorithm slightly outperforms the Random Forest algorithm, especially when combined with appropriate feature selection. Even three years after the training period, the results remain very good, with an <i>RMSE</i> of 0.59 m/s, <i>rRMSE</i> of 17%, and <i>R</i><sup><i>2</i></sup> of 0.84. The second stage assesses the multivariate approach's ability to forecast wind speed evolution at different time horizons (1–12 h) during a week characterized by significant wind dynamics. The forecasts demonstrate excellent agreement with observations at a 1-h time horizon, with an <i>RMSE</i> of 0.35 m/s, <i>rRMSE</i> of 7.6%, and <i>R</i><sup><i>2</i></sup> of 0.98, surpassing or comparable to other literature results. However, as the time lag increases, the <i>RMSE</i> (0.86, 1.14, and 1.51 m/s for 3, 6, and 12 h, respectively) and <i>rRMSE</i> (18.7%, 24.8%, and 32.9% for 3, 6, and 12 h, respectively) also increase, while <i>R</i><sup><i>2</i></sup> decreases (0.86, 0.79, and 0.60). Furthermore, the wind variations' amplitude is underestimated. To address this bias, a simple correction method is proposed.</p>\",\"PeriodicalId\":51132,\"journal\":{\"name\":\"Meteorology and Atmospheric Physics\",\"volume\":null,\"pages\":null},\"PeriodicalIF\":1.9000,\"publicationDate\":\"2023-11-20\",\"publicationTypes\":\"Journal Article\",\"fieldsOfStudy\":null,\"isOpenAccess\":false,\"openAccessPdf\":\"\",\"citationCount\":\"0\",\"resultStr\":null,\"platform\":\"Semanticscholar\",\"paperid\":null,\"PeriodicalName\":\"Meteorology and Atmospheric Physics\",\"FirstCategoryId\":\"89\",\"ListUrlMain\":\"https://doi.org/10.1007/s00703-023-00999-6\",\"RegionNum\":4,\"RegionCategory\":\"地球科学\",\"ArticlePicture\":[],\"TitleCN\":null,\"AbstractTextCN\":null,\"PMCID\":null,\"EPubDate\":\"\",\"PubModel\":\"\",\"JCR\":\"Q3\",\"JCRName\":\"METEOROLOGY & ATMOSPHERIC SCIENCES\",\"Score\":null,\"Total\":0}","platform":"Semanticscholar","paperid":null,"PeriodicalName":"Meteorology and Atmospheric Physics","FirstCategoryId":"89","ListUrlMain":"https://doi.org/10.1007/s00703-023-00999-6","RegionNum":4,"RegionCategory":"地球科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"Q3","JCRName":"METEOROLOGY & ATMOSPHERIC SCIENCES","Score":null,"Total":0}
Prediction and forecast of surface wind using ML tree-based algorithms
This study focuses on the importance of reliable surface wind forecasts for various sectors, particularly energy production. Traditional numerical weather prediction models are facing limitations and increasing complexity, leading to the development of machine learning models as alternatives or supplements. The research consists of two stages. In the first stage, the ERA5 database is used to evaluate the long-term performance of different combinations of features and two tree-based algorithms for predicting surface wind characteristics (speed and direction) in Cairo. The XGBoost algorithm slightly outperforms the Random Forest algorithm, especially when combined with appropriate feature selection. Even three years after the training period, the results remain very good, with an RMSE of 0.59 m/s, rRMSE of 17%, and R2 of 0.84. The second stage assesses the multivariate approach's ability to forecast wind speed evolution at different time horizons (1–12 h) during a week characterized by significant wind dynamics. The forecasts demonstrate excellent agreement with observations at a 1-h time horizon, with an RMSE of 0.35 m/s, rRMSE of 7.6%, and R2 of 0.98, surpassing or comparable to other literature results. However, as the time lag increases, the RMSE (0.86, 1.14, and 1.51 m/s for 3, 6, and 12 h, respectively) and rRMSE (18.7%, 24.8%, and 32.9% for 3, 6, and 12 h, respectively) also increase, while R2 decreases (0.86, 0.79, and 0.60). Furthermore, the wind variations' amplitude is underestimated. To address this bias, a simple correction method is proposed.
期刊介绍:
Meteorology and Atmospheric Physics accepts original research papers for publication following the recommendations of a review panel. The emphasis lies with the following topic areas:
- atmospheric dynamics and general circulation;
- synoptic meteorology;
- weather systems in specific regions, such as the tropics, the polar caps, the oceans;
- atmospheric energetics;
- numerical modeling and forecasting;
- physical and chemical processes in the atmosphere, including radiation, optical effects, electricity, and atmospheric turbulence and transport processes;
- mathematical and statistical techniques applied to meteorological data sets
Meteorology and Atmospheric Physics discusses physical and chemical processes - in both clear and cloudy atmospheres - including radiation, optical and electrical effects, precipitation and cloud microphysics.