Refining solar energy forecasting with optimization and feature engineering

Abstract

This research article addresses the imperative need for precise solar power generation forecasting to efficiently integrate solar energy into existing power grids. It introduces a holistic approach, considering multiple parameters and employing advanced modeling techniques. Emphasizing the importance of reliable raw input data, including solar irradiance, temperature, humidity, wind speed, and power generated, the study applies preprocessing methods such as data cleaning, outlier removal, and normalization for data integrity. A unique method for handling missing entries and feature extraction using LTF-MICF modeling captures essential characteristics. The HOpt_CLA-CBiGRU network model combines hierarchical optimization, CLA, and CBiGRU techniques for forecasting, while Mod_MUD optimizes hyperparameters. The proposed architecture demonstrates practical implementation, significantly contributing to forecasting accuracy and facilitating solar energy integration. Experimental results showcase superior performance in RMSE and MAE compared to baseline methods, highlighting the model's efficacy in supporting reliable power supply through optimal resource utilization and grid integration. The proposed strategy offers a foundation for efficient, sustainable energy management, heralding a future rich in green energy.