Global Horizontal Irradiance Estimation Using Bi-LSTM Algorithm

Accurate global horizontal irradiance (GHI) estimation is crucial for integrating renewable energy sources into the primary electrical system because it has an impact on grid operations, stability, and planning. Solar power is a significant component of the mix of energy sources since it has benefits for the environment and can reduce dependency on fossil fuels. With the aid of precise GHI projections, grid operators are able to forecast energy generation and adjust in real time to maintain a stable power supply. The planning of conventional power plants and efficient energy flow management is impacted by accurate GHI estimates. The balance between power production and consumption is essential for the stability of the grid. This research explores the potential of deep learning algorithms such as unilateral-LSTM and bilateral-LSTM networks for GHI prediction in Rajasthan, India, utilizing their inherent memory and adaptability. The min–max scaler technique is applied for preprocessing datasets obtained from the National Solar Radiation Database (NSRDB). To check the predictive performance and limitations of each model, this study analyzes various statistical parameters such as RMSE, MAE, and R2. The Bi-LSTM model using “RELU” as a gate activation function outperforms the Uni-LSTM due to getting the lowest values of MAE and RMSE. The highest R2 value obtained by the Bi-LSTM model indicates that its predictive performance is better than Uni-LSTM.