Key points are not available for this paper at this time.
In this paper, the performance of different technology photovoltaic (PV) systems was modeled using autoregressive integrated moving average (ARIMA) processes. Measurements from mono-crystalline (mono-c-Si), multi-crystalline (multi-c-Si) and amorphous (a-Si) silicon, cadmium telluride (CdTe) and copper indium gallium diselenide (CIGS) systems were used to construct monthly dc performance ratio (PR) time-series, from outdoor measurements. Each PR time-series was modeled a) with multiplicative ARIMA, b) with linear regression and c) with Seasonal-Trend Decomposition by Loess (STL) using the first 4 years of each time-series in order to compare the accuracy of the different methods. The models were used to forecast the PR of the 5 th year of the different PV technologies and the results from the aforementioned statistical methods were compared based on the root-mean-square error (RMSE). The results showed that ARIMA produced the lowest RMSE for crystalline silicon (c-Si) technologies, whereas for thin-film technologies, STL was more accurate. The results from ARIMA also showed that thin-film technologies were optimally modeled with identical model orders, whereas for c-Si, each technology required a different optimal model order.
Phinikarides et al. (Sat,) studied this question.