研究目的
To determine the capacity of the newly initiated hybrid FOS-ELM for predicting global solar radiation in the Burkina Faso region, with specific contributions including applying BIC for variable selection, proposing a novel hybrid AI model, adding regularization and forgetting factors to ELM, and validating against classical ELM.
研究成果
The FOS-ELM model demonstrated superior prediction accuracy for global solar radiation compared to classical ELM, with significant improvements in RMSE and MAE (68.8–79.8%). The model effectively handles non-stationary time series and requires no pre-specific knowledge of control parameters, making it a robust tool for renewable energy applications.
研究不足
The study is limited to a specific region (Burkina Faso) and uses daily data; future work could optimize network structure, include more meteorological factors, and apply to other regions or time scales.
1：Experimental Design and Method Selection:
The study uses a regularized online sequential extreme learning machine with variable forgetting factor (FOS-ELM) for solar radiation prediction, based on long-term daily data. Bayesian Information Criterion (BIC) is applied for input variable selection.
2：Sample Selection and Data Sources:
Historical daily meteorological data from Bur Dedougou, Burkina Faso, covering 15 years (1 January 1998–31 December 2012), with 4977 days for training and 500 observations for testing. Input variables include wind speed (Wspeed), maximum and minimum temperature (Tmax, Tmin), maximum and minimum humidity (Hmax, Hmin), evaporation (Eo), and vapor pressure deficiency (VPD).
3：List of Experimental Equipment and Materials:
Meteorological station data; Matlab software (R2017b, MathWorks) for algorithm implementation.
4：Experimental Procedures and Operational Workflow:
Data collection, BIC-based input combination selection, FOS-ELM model training and testing, daily prediction and parameter updates, statistical evaluation.
5：Data Analysis Methods:
Statistical indicators such as correlation coefficient (r), root mean square error (RMSE), mean absolute error (MAE), Willmott's Index (WI), Nash-Sutcliffe efficiency (NSE), Legates and McCabe's agreement (LM), relative RMSE (RRMSE), and relative MAE (RMAE) are computed for accuracy assessment.
独家科研数据包，助您复现前沿成果，加速创新突破
获取完整内容
