The integration of renewable energy resources into microgrid (MG) systems has significantly increased system non-linearity, leading to challenges such as voltage instability and higher energy costs. This paper introduces a neural network predictive controller (NNPC) as an innovative solution for minimizing voltage deviations and optimizing energy costs in dynamic MG environments. The proposed NNPC leverages machine learning (ML) techniques to forecast voltage deviations and provide real-time control signals to smart inverters, ensuring system stability and efficiency. The system employs a hierarchical control strategy, integrating droop control mechanisms and adaptive feedback to address the complexities of non-linear MG operations. Simulations conducted in MATLAB demonstrated that the NNPC outperformed traditional controllers, such as proportional integral derivative, proportional integral and proportional derivative controllers, in maintaining voltage stability. The NNPC achieved a mean voltage deviation of 257.5 V with a transient response time of 0.0026 s, showcasing superior adaptability to dynamic MG conditions. This work highlights the transformative role of artificial intelligence in enhancing MG performance, offering a scalable and efficient solution for modern energy systems.