Smart inverters are essential for the effective conversion and distribution of electricity in contemporary power systems. However, the increased reliance on smart grid technologies makes these systems vulnerable to cybersecurity risks, including False Data Injection Attacks (FDIAs). By manipulating data and deceiving control systems, these attacks have the potential to compromise grid stability. To enhance the security and resilience of smart inverters against FDIAs, this article proposes an innovative deep learning-based system. By identifying anomalous data patterns through deep learning models, the proposed approach enables the rapid detection and blocking of FDIA attempts. The framework is designed to generalize across different operating environments by training the model with a diverse dataset of normal and attack events, ensuring reliable detection even in complex scenarios. This method offers a viable solution to protect smart inverters and maintain the stability of the power grid in the face of evolving cyber threats.