Machine Learning-Based Estimation of Output Current Ripple in PFC-IBC Used in Battery Charger of Electrical Vehicles: A Comparison of LR, RF and ANN Techniques

In this study, an artificial neural network (ANN) model is developed for the purpose of estimating the output current ripple of a power factor correction (PFC) AC/DC interleaved boost converter (IBC) used in battery charger of electrical vehicles (EVs) based on the inductance current ripple, switching frequency and load changes. Besides, the improved ANN model is compared with some different machine learning (ML) techniques like linear regression (LR), random forest (RF). The PFC-IBC is simulated with the PSIM simulation program to estimate the output current ripple. As a result, 336 output current ripple values are obtained based on inductance current ripple, different switching frequency and load changes. Then, the value of output current ripple is estimated by training the input parameters with LR, RF and ANN machine learning techniques (MLTs) for controlling the current harmonics drawn from the grid and for reliable charging of batteries. It is seen that the estimation value obtained with MLTs is quite compatible with the actual value obtained with the simulation. In addition, in the study carried out with the simulation, it takes a period of several days to obtain the estimation results; whereas, the operation of estimation with MLTs can be completed in a short period such as a few minutes. This clearly reveals the advantage of the MLTs. Therefore, this value is estimated through the MLTs with a high accuracy before the design of the charging device in order to maintain at a secure level the output current ripple posing considerable importance in electrical vehicle battery charge. Also, in this estimation process, LR, RF and developed ANN techniques are examined and compared separately in the WEKA program and it is observed that the developed ANN model proposes better results than other techniques.