A GRID SEARCH OPTIMIZED HYBRID MODEL FOR SOLAR IRRADIANCE FORECAST USING EMPIRICAL MODE DECOMPOSITION AND BIDIRECTIONAL GATED RECURRENT UNIT

To ensure the reliable operation of the power grid, an accurate forecasting of solar irradiance (SI) is highly essential. Classical forecasting methods struggle to capture the fundamental nonlinearity of SI, resulting in low forecasting accuracy. A hybrid model based on deep learning (DL) technique called EMD-GS-BiGRU, which is the combination of empirical mode decomposition (EMD) with Bidirectional gated recurrent unit (BiGRU) predictor and grid search algorithm is proposed in this paper. The suggested EMD-GS-BiGRU model constructed and tested in two different Indian locations: Chennai and Jammu, for different popular models: artificial neural network (ANN), long short-term memory (LSTM), gated recurrent unit (GRU), and their EMD-based hybrid version i.e. EMD-GS-LSTM and EMD-GS-GRU. The evaluation of the performance of the proposed model is performed by statistical and graphical analysis of one hour ahead seasonal SI forecasts. According to the findings, the lower average root means square error (RMSE) of 64.39 W/m2-83.46 W/m2 is obtained by proposed model compared to contrast models. The proposed model also obtained higher yearly R2 value, 0.9366-0.955 compared to other models. Furthermore, the study found that the EMD with grid search optimized BiGRU enhanced the RMSE (39.23%-47.31%) and mean absolute percentage error (MAPE) (41.82%-52.31%) compared to other standalone contrast models.