Neural Network Modeling and Simulation of A 265W Photovoltaic Array

Full Text PDF PDF
Author(s) Guda, H. A | Aliyu U. O.
Pages 113-125
Volume 5
Issue 3
Date March, 2015
Keywords Solar Cell, Pv Array, Neural Network, Radial Basis Function, Simulation, Absolute Error


This paper presents the Neural Network modeling and simulation of a 265 Watts photovoltaic array installed at the Faculty of Engineering and Engineering Technology of Abubakar Tafawa Balewa University, Bauchi, Nigeria. Hitherto, Mathematical modeling is the favoured method for characterizing photovoltaic (PV) arrays. This approach would require detailed information on the physical parameters relating to the solar cell material, which may not be readily available. Even in situations where the required information is provided on the manufacturer’s datasheet, it tends not to be very accurate as it is not representative of the actual field performance of the array. Thus results obtained from mathematical modeling of photovoltaic arrays are only accurate to the extent of the accuracy of the model parameters. A better PV array characterization approach is to use Neural Network modeling because it does not require any physical definitions of the array and hence has the potential to provide a superior method of characterization than the already established conventional techniques. In this paper, two Radial Basis Function Neural Network (RBFNN) trained models are employed to simulate the performance of a 265 Watts photovoltaic array. The first model predicts the array I-V and P-V curves while the second predicts its maximum power for all operating weather conditions. Results of array performance plots show close correlation with those obtained through conventional mathematical modeling. RBFNN returned absolute errors of 1.794 %, 1.594 % and 1.262 % with respect to PV maximum power predictions for harmattan, cloudy and clear sunny seasons respectively.

< Back to March Issue