Modeling wind power uncertainty in the long-term operational reserve adequacy assessment: A comparative analysis between the Naïve and the ARIMA forecasting models

The growing integration of renewable energy in power systems demands for adequate planning of generation systems not only to meet long-term capacity requirements but also to cope with sudden capacity shortages that can occur during system operation. As a matter of fact, system operators must schedule an adequate amount of operational reserve to avoid capacity deficits which can be caused by, for instance, overestimating the wind power that will be available. The framework proposed for the long-term assessment of operational reserve relies on the Naïve forecasting method to produce wind power forecasts for the next hour. This forecasting model is simple and widely used to obtain short-term forecasts. However, it has been shown that regression models, such as the Autoregressive Integrated Moving Average (ARIMA) model, can outperform the Naïve model even for forecasting horizons of up to 1 hour. This paper investigates the differences in the risk indices obtained for the long-term operational reserve when using the Naïve and the ARIMA forecasting models. The objective is to assess the impact of the forecasting error in the long-term operational reserve risk indices. Experiments using the Sequential Monte Carlo Simulation (SMCS) method were carried out on a modified version of the IEEE RTS 79 test system that includes wind and hydro power variability. A sensitivity analysis was also performed taking into account several wind power integration scenarios and two different merit orders for scheduling generating units.