ANN-based estimation of the voltage ripple according to the load variation of battery chargers

In this study, the low DC voltage obtained from the fuel cell (FC) is boosted in the output voltage range of 500V by using a double cascaded boost converter (DCBC) circuit. The power electronic circuit has been simulated with the parametric analysis and there have been obtained output voltage ripple about 1295 values at the converter output according to the load variation and the different switching frequencies. After that, the output voltage ripple has estimated with Artificial Neural Network (ANN) for adaptive frequency control depending on the load variation. Thus, the output voltage ripple is taught to the system according to the non-linear load change and it can be kept at a safe level with the frequency control. In such a circuit, the parametric simulation time takes about a week according to the computational performance. The ANN model eliminates the mathematical procedures and time consumption for circuit simulations. The results obtained with the ANN model are in harmony with the simulation results, and these results show that it can be successfully used to estimate the output voltage ripple. In this way, the ANN model has been developed with very high accuracy (<0.5%) in that the charging voltage ripple of Li-Ion batteries can be kept at safe levels.