Enhancing Solar PV Power Generation Forecasting: An Ensemble Approach with LSTM and Prophet Models

Smart grid initiatives have greatly increased the share of renewable energy in the grid, highlighting the challenges of managing its irregular and unpredictable output, which makes accurate forecasting crucial for aligning future renewable production with demand. Several methods have been presented by researchers as discussed in the introduction section. With non-seasonal patterns and constantly changing environmental conditions, it is very important to address the forecasting issues.Realizing that a single model cannot generate reliable forecasts for solar PV power generation. The proposed method leverages the LSTM model's ability to capture long-term dependencies and non-linear patterns, alongside the Prophet model's strength in handling seasonality and external regressors such as holidays and maintenance schedules.The results indicate that the Proposed Model outperforms all other models, achieving the highest R<sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sup> of 98.4% (plant site) and 94.68% (website) for hourly sampled data, along with the lowest RMSE values of 15.9 and 11.03, respectively. For daily average data, the Proposed Model maintains an R<sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sup> of 93.4% across both datasets, with RMSE values of 5.78 (plant site) and 6.92 (website). Among LSTM variants, LSTM with Layer Normalization demonstrated the best improvements, increasing R<sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sup> to 92.46% while reducing RMSE. Prophet models, particularly Prophet with Weather Regressor, showed moderate improvements, but Prophet with Smoothing did not significantly enhance forecasting accuracy.Findings verify that the proposed ensemble model consistently outperforms individual models (LSTM/Prophet), even when data are collected at shorter intervals. This integrated strategy improves forecasting accuracy, which helps energy authorities make well-informed decisions for efficient energy distribution planning.