With the increasing significance of renewable energy in addressing climate change, it has become crucial to have precise predictions for short-term solar photovoltaic electricity production. The existing approaches, such as Random Forest, Decision Tree, LightGBM and XGBoost, have limits in terms of accuracy and dependability when it comes to predicting solar output. This study presents a new hybrid ensemble learning model that combines LightGBM, XGBoost and Random Forest. The model’s hyperparameters are optimised using Grid Search. The model exhibits remarkable precision and resilience, outperforming current techniques in two separate power plants. Significantly, it provides flexibility in various environmental situations, demonstrating its promise as a reliable solution for predicting short-term solar PV AC power. The paper makes a substantial contribution to the field of predictive modelling in renewable energy by providing valuable insights into alternative ensemble approaches and the integration of real-time data for dynamic forecasting scenarios. This study facilitates improved decision-making in the management of renewable energy, hence advancing the incorporation of solar electricity into the electrical grid.