Investigation of ANN performance for tracking the optimum points of PV module under partially shaded conditions

Abstract

Solving partially shaded condition still remains an important task in the PV system practice. Under such condition, the global maximum point shifts continuously in wide voltage range from the local maxima, which make it difficult for conventional controllers. Intelligent techniques based on artificial neural network are well-known as the promising methods to identify the global maxima. However, there are many variants of neural networks and they have strong and weak points during the implementation. This paper investigates the performance of radial basis function (RBF) neural network and three layered feed-forward neural network (TFFN) under partial shadow operation of PV module. These two ANN structures are well-recognized for optimization, forecasting and control in PV system application due to the simplicity of structure and high performance accuracy. The investigation is focused on the network structure, training and validation process of these methods. To determine to which method is preferable to handle this task, the adaptive neuro-fuzzy inference system (ANFIS) is used as the comparator. The proposed method is verified and tested using developed real-time simulator.

Access to Document

Other files and links

• Link to publication in Scopus

Fingerprint