New Decision Tree Controller for MPPT Based on Fuzzy Logic Controller Data

Abstract

One of the main parts of the grid-connected PV system is the DC/DC boost converter which several research studies aim to improve. This component is commanded by a control algorithm which allows it to identify and track the maximum power point of the PV array throughout any weather conditions. Several approaches in the literature are proposed to accomplish this task, among the well-known, the intelligent commands such as Fuzzy Logic Controller (FLC). although FLC has many advantages such as convergence time, precision, efficiency, etc. there are always some drawbacks like the complexity of the controller, the hard algorithm implementation. For this purpose, a new approach has been proposed, using the FLC dataset retrieved from the simulation to build a decision tree model (DTM) obtaining both advantages of FL and DTM. Therefore, 66% of the data set will be allocated to the learning process and the rest is for the model test. The learning process is performed based on the C4.5 algorithm, where splitting criteria used to form the tree is called the gain ratio, this one is applied over the data set recursively to find the best model that fittest with FLC data. A validation step was carried out to verify the performances of the model by using new data unseen by the learning process. Lastly, the results have shown that the DT model has a high correlation with the FLC combining both advantages of the two approaches, moreover, the new model has compared with P&O and PID controllers and shown high effectiveness.