A Connectionist Based Approach For Reducing The Number Of Voltage Sensors In Modular Multilevel Converter

Intensive research is being directed to renewable energy due to their significant features. The converter is the core of the renewable energy, system. Recently, a converter topology called Modular Multilevel Converter (MMC) has been introduced whose main challenge is the balance of its Submodules (SMs) capacitors voltages. Hence, a voltage sensor is required for each SM to measure its capacitor voltage; which increase the system cost. Recently, soft computing techniques have been successfully applied as regression tools to deal with many challenges characterized by complex dynamical representations. In this paper, a soft computing approach based on connectionist modeling (artificial neural networks) is used to predict the SM capacitor voltages. Moreover, each arm employs only one voltage sensor to reduce the system cost; whose function is to measure the output voltage of a set of a series connected SMs and updates the predictors when there is only one SM activated within the set.