Title: Optimum ANN empirical model of capacitive deionisation desalination unit

Authors: Adel El Shahat; Rami J. Haddad; Youakim Kalaani

Addresses: Department of Electrical Engineering, Allen E. Paulson College of Engineering and Information Technology, Georgia Southern University, GA, USA; Faculty of Petroleum and Mining Engineering, Suez University, Egypt ' Department of Electrical Engineering, Allen E. Paulson College of Engineering and Information Technology, Georgia Southern University, GA, USA ' Department of Electrical Engineering, Allen E. Paulson College of Engineering and Information Technology, Georgia Southern University, GA, USA

Abstract: Capacitive deionisation (CDI) has emerged as a robust energy efficient for water desalination. In this paper, a novel CDI electrosorption process is proposed to increase the efficiency based on real experimental data. It is achieved by artificial neural network (ANN) to develop four models. For problem formulation, closed forms mathematical equations were derived, thus, resulting in a very efficient programming algorithm. Optimum patterns ANN models were validated by implementing two ANN units to drive the CDI electrosorption process. This proposed method was tested and verified using actual and predicted ANN values which yielded excellent results with regression factors between 0.99983 to 1. Optimum patterns are validated in the form characteristics comparisons between genetic and original one. The ANN models their algebraic equations are adopted for various characteristics estimation process. They created with suitable numbers of layers and neurons that provided fast and accurate network training.

Keywords: capacitive deionisation; CDI; modelling; artifical neural networks; ANNs; genetic algorithms; optimisation; estimation; water desalination; energy efficiency; electrosorption.

DOI: 10.1504/IJIED.2015.069785

International Journal of Industrial Electronics and Drives, 2015 Vol.2 No.2, pp.116 - 133

Received: 26 Feb 2014
Accepted: 16 Dec 2014

Published online: 11 Jun 2015 *

Full-text access for editors Access for subscribers Purchase this article Comment on this article