Modern power grids are becoming increasingly complex with the integration of renewable energy sources, decentralized generation, and smart technologies, which pose significant challenges to reliability and operational efficiency. Traditional fault detection and maintenance methods, often reactive and rule-based, suffer from delayed fault identification, high false alarm rates, and inefficient maintenance scheduling. This paper proposes a hybrid artificial intelligence (AI) framework that integrates machine learning (ML) models-such as decision trees, support vector machines, and neural networks-for fault detection with optimization algorithms based on genetic algorithms for predictive maintenance scheduling. The hybrid approach enables both accurate fault diagnosis and cost-effective maintenance planning within a unified framework. Simulation studies conducted using real-world power grid sensor data demonstrate that the proposed method achieves a fault detection accuracy of 92.5 %, reduces maintenance costs by 30 %, and decreases system downtime by 25 % compared to traditional rule-based and standalone AI approaches. These results confirm the effectiveness of the hybrid AI model in enhancing grid reliability, reducing operational costs, and enabling proactive maintenance. The proposed framework provides a scalable, realtime solution for managing the increasing complexities of future smart grids.