In a previous work, a moisture diffusion coefficient for non-impregnated transformer insulation paper was proposed, which is useful for analysing the moisture dynamics in electrical power transformer insulation during its construction stage. This diffusion coefficient was determined using a set of experimental drying curves obtained in a thermogravimetric analyzer and an optimization process that employs genetic algorithms. The expression for the moisture diffusion coefficient proposed in that previous work is dependent on the temperature, insulation thickness and average moisture concentration. The use of average moisture concentration is not common in diffusion models. In this paper, a new expression for the moisture diffusion coefficient is proposed in which the dependence on average moisture has been substituted by a dependence on the local concentration. This new expression was obtained that is more suitable for diffusion models. To obtain that expression, the same experimental data used in the previous study have been used, but the genetic algorithm's optimization process was modified. In this paper, a description of the applied procedure is detailed, and the new proposed expression is presented. Finally, a comparison between both equations, the one dependent on average moisture and the one dependent on the local moisture, is performed using a drying diffusion model implemented in a finite element commercial software package. (C) 2012 Elsevier Ltd. All rights reserved.