Smart grid technology is revolutionizing the electrical industry by enabling an efficient, digitized energy ecosystem that allows smooth energy flow from production to end consumers. Traditional grid systems that are not smart enough to estimate the user's power requirements are unreliable. The concept of smart grids was therefore birthed from there, with smart technologies such as IoT, AI, and decentralized energy management platforms that aim at maximizing energy distribution and reducing transmission losses. This research introduced machine learning (ML) and deep learning (DL) models for smart grid stability prediction. Our proposed model, TabTransformer obtained the best results on the assessed methodologies at 99.40% test accuracy and an AUC-ROC score of 1.0. Additionally, our research combined with explainability approaches in the form of SHAP and LIME that will further provide insight into the feature contributions that give assurances of dependability and robustness. Our research results show the great potential artificial intelligence-based methodologies have on smart grids with sustainable and efficient energy systems. This research promotes the use of advancement to attain energy stability and resilience in the modern power environment.