Neural Network Based Model Predictive Controllers for Modular Multilevel Converters

Modular multilevel converter (MMC) has attracted much attention for years due to its good performance in harmonics reduction and efficiency improvement. Model predictive control (MPC) based controllers are widely adopted MMC because the design is straightforward different objectives can be simply implemented a cost function. However, computational burden of MPC imposes limitations implementation many possible switching states. To solve this, we machine learning (ML) on data collection from algorithm. The ML models trained emulate which effectively reduce computation real-time since built with simple math functions that not correlated complexity method applied this study neural network (NN) there two types establishing controllers: NN regression pattern recognition. Both using sampled tested system. A comparison experimental results shows better lower than