A study on the lifetime estimation of photovoltaic modules under accelerated environmental conditions

Lifetime estimation is critical for predicting the revenue of the photovoltaic (PV) plant to be built. Therefore it must be done before huge investments will be made to build and operate a PV plant. Since the degradation rate of PV modules features as a random variable, statistical models are commonly used to estimate the lifetime of PV modules. However, most existing researches select the statistical model to describe the degradation rate of PV modules empirically, and lack of basis. In this paper we propose a practical procedure for selecting the statistical model which describes the degradation rate of PV modules best. Based on the selected model, the lifetime of PV modules is estimated. We use probability plots and hypothesis tests to validate the distributional assumptions. Then we use negative log-likelihood values to further determine the distribution that describes the lifetimes best. Three classical distributions, i.e., Weibull, lognormal and exponential distributions, are considered. The life distributions at different stress levels are assumed to come from the same parametric family. Based on the assumption, the parameters of the hypothesized distributions are estimated using maximum likelihood estimation methods. It is found that the PV lifetimes follow a parametric location-scale distribution family, and they follow the Weibull distribution best. The conclusion is realistic from a physical point of view.