Authors: Nilesh.P.Dabe, Yogesh R. Patni, Deepak Kadam, Kulkarni Kirti S
Abstract: Accurate electricity load forecasting is critical for maintaining stability, reliability, and cost efficiency in modern smart grids, especially with the growing integration of renewable energy sources. However, the inherent intermittency and uncertainty of renewables such as solar and wind introduce significant challenges for traditional forecasting models. This paper proposes a Hybrid Generative AI–Enhanced Load Forecasting Model that combines Generative Adversarial Networks (GANs) with deep learning architectures to improve prediction accuracy under varying renewable energy conditions. The generative component synthesizes high-variance energy patterns that capture extreme fluctuations, while the predictive module leverages a hybrid CNN–LSTM network for temporal–spatial learning. Experimental results on real-world datasets demonstrate substantial improvements, with reductions of 40.1% in MAE, 38.2% in RMSE, and enhanced robustness against high-uncertainty renewable inputs. The proposed model also reduces load–supply mismatch by 42.4% and energy imbalance cost by 41.3%, leading to more efficient power distribution and operational cost savings. These findings highlight the potential of Hybrid Generative AI to significantly enhance smart grid forecasting performance and support resilient, data-driven energy management strategies.
