Efficient Time Series Photovoltaic Production Forecasting using FLAML with Traditional Machine Learning Algorithms

This study explores the use of automatic photovoltaic production forecasting. The library is applied with five traditional machine learning algorithms, including Xgb_Limitdepth, Random-Forest, Extra-Tree, XGBoost, and lightGBM. The dataset used in this study is obtained from a photovoltaic power plant located in a sunny region in Europe, Greece, and it includes hourly power production data for a period of four years. The results show that FLAML can automatically optimize hyperparameters and select the best algorithm for the given data, achieving accurate forecasts for photovoltaic power production. The experiment shows that the combination of FLAML with traditional machine learning algorithms performs better than using each algorithm individually. In addition, the study compares the performance of the different algorithms and demonstrates that XGBoost and LGBM outperform the other algorithms. Overall, this study demonstrates the usefulness of FLAML in time series forecasting and its potential to optimize the forecasting process by automating the selection of the best algorithm and hyperparameters. This study’s empirical evidence can be useful for researchers and practitioners in the field of renewable energy, particularly in photovoltaic power production, and similar can be used for wind power production and tidal power production.