In the energy harvesting domain, the modelling of a solar panel plays an important role in predicting the energy availability of energy harvesting system applications. Indoor environments, which are illuminated by artificial light sources, have typically much lower illumination levels than outdoor environments. In this paper, we compare the behaviour of different types of models under low illuminance conditions, in order to investigate sufficient modelling approaches for indoor environments. Previous work has shown that equivalent circuit modelling may have reduced performance under low illuminance conditions. Instead, we investigate behavioural models and compare their results with the equivalent circuit model. Two different types of behavioural models have been tested, namely artificial neural network models and polynomial curve fitting models. The comparison of these three models has shown that it is not possible to establish which of the modelling methods performs best, because each of them have strong points and shortcomings making the ideal choice application dependant.