Neural network adaptive wavelets for sizing of stand-alone photovoltaic systems

Single layer feed-forward neural networks with hidden nodes an, adaptive wavelet functions have been successfully demonstrated to have potential in many applications. In this paper, an application to sizing of stand-alone PV systems design method of an unknown optimal sizing combination is presented. These optimal sizing combinations allow to the users of stand-alone PV systems to determine the number of solar panel modules and storage batteries necessary to satisfy a given consumption, especially in isolated sites where the global solar radiation data is not always available. A developed model combine between Multilayer percepron (MLP) and Infinite Impulse Filter (FIR), this IIR recurrent structures is combined by cascading to the network to provide double locale structure resulting in improving speed of learning. The MLP-IIR model has been trained by using 200 known sizing combinations data corresponding to 200 locations. In this way, the adaptive model was trained to accept and even handle a number of unusual cases. Known sizing coefficients were subsequently used to investigate the accuracy of estimation. The, training MLP-IIR model was performed with adequate accuracy. Subsequently, the unknown validation sizing coefficients set produced very set accurate estimation with the correlation coefficient between the actual and the MLP-IIR model estimated data of 98% was obtained. This result indicates that the proposed method can be successfully used for estimating of optimal sizing combinations of stand-alone PV systems for any locations in Algeria, but the methodology can be generalized using different locations in the world. Also, obtained results by feed-forward (MLP), Radial basis function (RBF) and an adaptive MLP-IIR model have been compared with measured data in order to illustrate the importance of the new developed model. Possible application can be found in: rural sites, pumping water, electrification in isolated sites.