Predictability Index for Renewable Energy and Uncertainty Quantification with Analog Ensemble

Abstract:

Renewable energy, in forms of wind and solar, is significantly affected by weather. For example, solar energy production depends on factors like ambient temperature and cloud formation. Predictions of the actual power generation are generally uncertain, primarily because of the uncertainty associated with weather forecasts. However, in addition to the uncertainty associated with weather, changes in weather patterns, more frequent extreme and rare events, and conditions of the power facilities increase the difficulty of making accurate predictions. This is challenging when integrating such a form of renewable energy source to the operation power grid as the power generation is not fully controlled and its high variability increases the operational cost of the power grid.

In this research, we present that the predictability of an energy source should be treated separately from the internal variability. For example, solar energy can still be integrated as long as it is predictable and the uncertainty level is controlled, even if there is a large variation. The predictability of the renewable source should be considered during the decision making process. Currently, however, the main consideration still remains to be the amount of annual average irradiance in the case of photovoltaic solar and long-term variation.

We propose to couple Analog Ensemble technique (AnEn) and a solar energy production simulator to study the predictability and uncertainty of photovoltaic solar energy production over the CONUS. Figure a shows the predicted annual power generation for 2018 from Analog Ensemble mean and figure b shows the hourly-averaged standard deviation of the ensemble members. The vastly different spatial pattern suggests that it is not sufficient to only consider the amount of energy available, but also to consider the predictability and uncertainty. A predictability index should be formulated to consider different aspects of the renewable energy source of interest. The goal of this work is to gain insights into how predictable different regions in the US are and how to optimize power generation using different types of panel configuration.