In this thesis, the use of predictive control methods based on artificial intelligence, machine learning, and artificial neural networks is investigated to solve modular multilevel converters' challenges. In modular multilevel converters, minimizing the circulating current and rippling the capacitors' voltage, controlling the load side current, and maintaining the capacitors' voltage balance is among the most crucial research goals of researchers in recent years. Therefore, the controller role used in such converters is vital. Today, with the increase in processor speed, it is possible to use new control methods such as predictive control. Using the system's dynamic model and predicting the future behavior of the variables, control is performed, and its implementation in converters is very easy. in predictive control of the MMC converter, researchers have proposed various techniques that have good results in these cases. However, the submitted proposals have a sizeable computational load, which should lead to the need to improve the predictive controller's performance to reduce the computational load and enhance the MMC converter challenges' performance. Among other things, the use of many cells in the MMC converter, in some .applications such as HVDC, increases the switching frequency and losses and the high volume of calculations in the controller. To solve this problem, a method must be provided to reduce the switching modes and, consequently, the computational load First, by expressing the advantages of predictive control methods and focusing on the FCS-MPC method, the general relationships related to the control of electronic power converters and then the challenges of modular multilevel converters using this control method are discussed. Then, by expressing the main challenges of this control method such as the time-consuming method of determining the weight factor, dependence on the converter circuit model, high volume of required calculations, switching losses, and high switching frequency, two proposed methods based on machine learning and artificial neural networks to improve this challenge Are similar to the FCS-MPC method, while maintaining the steady-state and dynamic performance of the converter. Next, focusing on modular three-level and five-level converters, these control methods are proposed to achieve two general goals.The first goal is to solve the challenges of modular multilevel converters. The second goal is to solve the FCS-MPC method's inherent difficulties and improve system performance in applying this control method. In the following, the results of the simulation of the two proposed ways and the FCS-MPC method are compared with each other, and the priority of each technique in terms of different performance characteristics is evaluated. Finally, conclusions and suggestions for further research are provided. Modular Multilevel Converter 2- Control Challenges of Modular Multilevel Converter 3- Predictive Control in Power Electronic 4- Machine Learning and Artificial Neural Networks