Application of genetic algorithms for the extraction of electrical parameters of multicrystalline silicon

Abstract

The application of the genetic algorithms (GAs) concept in measurements science deals with a fitting procedure used for the numerical prediction of physical parameters from an experimental data curve. In this work, GAs were applied for the extraction of electrical parameters of multicrystalline silicon solar cells. The experimental technique used is the light-beam-induced-current (LBIC). From LBIC measurements, we deduced the values of the diffusion length L and the grain boundary recombination velocity Vr of the minority carriers of the multicrystalline silicon wafers using the fitting procedure. The nonlinear fitting procedure is based on the minimization of the standard deviation of the theoretical LBIC profile from the experimental one. However, this criterion is not convex, and using traditional deterministic optimization algorithms leads to local minima solutions. To overcome this problem, the nonlinear least-square minimization technique was computed with the GAs strategy, increasing the probability of obtaining the best minimum value of the cost function in very reasonable time. The results of the proposed algorithm are presented and compared with the Levenberg–Marquardt algorithm and the Gauss–Newton method. The GAs-based numerical technique was found to be a promising and a powerful technique for numerical evaluation of the electrical parameters of silicon solar cells.