{"title":"预测太阳辐照度,为太阳能系统设计和融资决策提供依据","authors":"Ronewa Mabodi;Jahvaid Hammujuddy","doi":"10.23919/SAIEE.2024.10551303","DOIUrl":null,"url":null,"abstract":"This research presents the implementation and evaluation of machine learning models to predict solar irradiance (W/m\n<sup>2</sup>\n). The objective is to provide valuable insights for making informed decisions regarding solar system design and financing. A thorough exploratory data analysis was conducted on the Southern African Universities Radiometric Network (SAURAN) data collected at the University of Pretoria’s station to gain insights into the patterns of solar irradiance over the past 10 years. Python’s functions and libraries are utilized extensively for conducting exploratory data analysis, model implementation, model testing, forecasting, and data visualization. Random Forest (RF), k-Nearest Neighbors (KNN), Feedforward Neural Network (FFNN), Support Vector Regression (SVR), and eXtreme Gradient Boosting models (XGBoost) are implemented and evaluated. The KNN model was found to be superior achieving a relative Root Mean Squared Error (RMSE), relative Mean Absolute Error (MAE), and R-Squared (R\n<sup>2</sup>\n) of 5.77%, 4.51% and 0.89 respectively on testing data. The variable importance analysis revealed that temperature (X!) exerted the greatest influence on predicting solar irradiance, accounting for 44% of the predictive power. The KNN model is suitable to inform solar systems design and financing decisions. Directions for future studies are identified and suggestions for areas of exploration are provided to contribute to the advancement of solar irradiance predictions.","PeriodicalId":42493,"journal":{"name":"SAIEE Africa Research Journal","volume":null,"pages":null},"PeriodicalIF":1.0000,"publicationDate":"2024-06-06","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=10551303","citationCount":"0","resultStr":"{\"title\":\"Solar Irradiance Forecasting for Informed Solar Systems Design and Financing Decisions\",\"authors\":\"Ronewa Mabodi;Jahvaid Hammujuddy\",\"doi\":\"10.23919/SAIEE.2024.10551303\",\"DOIUrl\":null,\"url\":null,\"abstract\":\"This research presents the implementation and evaluation of machine learning models to predict solar irradiance (W/m\\n<sup>2</sup>\\n). The objective is to provide valuable insights for making informed decisions regarding solar system design and financing. A thorough exploratory data analysis was conducted on the Southern African Universities Radiometric Network (SAURAN) data collected at the University of Pretoria’s station to gain insights into the patterns of solar irradiance over the past 10 years. Python’s functions and libraries are utilized extensively for conducting exploratory data analysis, model implementation, model testing, forecasting, and data visualization. Random Forest (RF), k-Nearest Neighbors (KNN), Feedforward Neural Network (FFNN), Support Vector Regression (SVR), and eXtreme Gradient Boosting models (XGBoost) are implemented and evaluated. The KNN model was found to be superior achieving a relative Root Mean Squared Error (RMSE), relative Mean Absolute Error (MAE), and R-Squared (R\\n<sup>2</sup>\\n) of 5.77%, 4.51% and 0.89 respectively on testing data. The variable importance analysis revealed that temperature (X!) exerted the greatest influence on predicting solar irradiance, accounting for 44% of the predictive power. The KNN model is suitable to inform solar systems design and financing decisions. Directions for future studies are identified and suggestions for areas of exploration are provided to contribute to the advancement of solar irradiance predictions.\",\"PeriodicalId\":42493,\"journal\":{\"name\":\"SAIEE Africa Research Journal\",\"volume\":null,\"pages\":null},\"PeriodicalIF\":1.0000,\"publicationDate\":\"2024-06-06\",\"publicationTypes\":\"Journal Article\",\"fieldsOfStudy\":null,\"isOpenAccess\":false,\"openAccessPdf\":\"https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=10551303\",\"citationCount\":\"0\",\"resultStr\":null,\"platform\":\"Semanticscholar\",\"paperid\":null,\"PeriodicalName\":\"SAIEE Africa Research Journal\",\"FirstCategoryId\":\"1085\",\"ListUrlMain\":\"https://ieeexplore.ieee.org/document/10551303/\",\"RegionNum\":0,\"RegionCategory\":null,\"ArticlePicture\":[],\"TitleCN\":null,\"AbstractTextCN\":null,\"PMCID\":null,\"EPubDate\":\"\",\"PubModel\":\"\",\"JCR\":\"Q4\",\"JCRName\":\"ENGINEERING, ELECTRICAL & ELECTRONIC\",\"Score\":null,\"Total\":0}","platform":"Semanticscholar","paperid":null,"PeriodicalName":"SAIEE Africa Research Journal","FirstCategoryId":"1085","ListUrlMain":"https://ieeexplore.ieee.org/document/10551303/","RegionNum":0,"RegionCategory":null,"ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"Q4","JCRName":"ENGINEERING, ELECTRICAL & ELECTRONIC","Score":null,"Total":0}
Solar Irradiance Forecasting for Informed Solar Systems Design and Financing Decisions
This research presents the implementation and evaluation of machine learning models to predict solar irradiance (W/m
2
). The objective is to provide valuable insights for making informed decisions regarding solar system design and financing. A thorough exploratory data analysis was conducted on the Southern African Universities Radiometric Network (SAURAN) data collected at the University of Pretoria’s station to gain insights into the patterns of solar irradiance over the past 10 years. Python’s functions and libraries are utilized extensively for conducting exploratory data analysis, model implementation, model testing, forecasting, and data visualization. Random Forest (RF), k-Nearest Neighbors (KNN), Feedforward Neural Network (FFNN), Support Vector Regression (SVR), and eXtreme Gradient Boosting models (XGBoost) are implemented and evaluated. The KNN model was found to be superior achieving a relative Root Mean Squared Error (RMSE), relative Mean Absolute Error (MAE), and R-Squared (R
2
) of 5.77%, 4.51% and 0.89 respectively on testing data. The variable importance analysis revealed that temperature (X!) exerted the greatest influence on predicting solar irradiance, accounting for 44% of the predictive power. The KNN model is suitable to inform solar systems design and financing decisions. Directions for future studies are identified and suggestions for areas of exploration are provided to contribute to the advancement of solar irradiance predictions.