Neural Network-Based Sensorless Control of Flyback Converter for Cell Balancing

A conventional control system with an active clamp flyback converter uses a feedforward method that involves feeding the input to a mathematical model beforehand to adjust the duty cycle accordingly. The feedforward method uses a mathematical model that involves a lot of calculations to obtain an efficiency alike feedback system. Integrating a neural network trained with a feedback system data output can be used as a replacement for the mathematical model to have performance as par with the feedback system. Sensors are used to measure the voltages for the feedback circuit to work. This work investigates the potential of utilizing a neural network to enable sensorless operation within a feedforward control system to regulate the voltage input to the cell balancina system.