Intra-day Solar Resource Variability Model Application for a Day Ahead Forecast Uncertainty Prediction

Abstract —We introduce a new way of characterizing solar resource variability through a metric that we call Nominal Variability (NMV). The NMV is defined as a standard deviation of the absolute value of finite difference of the ratio of Global Horizontal Irradiance (GHI) to that of Clear Sky Irradiance (CSI) values. NMV has a polynomial relationship of the type concave downward when compared to the daily clear sky index (The sum of hourly GHI/ CSI). The inflection point of the polynomial is at the period when there is the highest solar resource variability for a day. The NMV gives a very valuable information on the resource variability at a given site, which will be a tool for risk assessment of long term energy production estimates. We have demonstrated the use of the NMV model to predict forecast uncertainty at both sub hourly and day ahead forecast time horizons. We have also developed NMV model for several locations using both ground and satellite data, the result shows that it is possible to use satellite data for areas where ground data are not available.

Index Terms — Satellite Data, Solar Forecast, Solar Resource, intra-day Solar Resource variability, Day Ahead Solar Forecast Uncertainty I. Introduction

The characterization of the solar resource is often calculated in terms of magnitude, that is how much solar energy is available at an area of interest over a given time. However, a complete characterization of solar resource should include the temporal and spatial variability the resource.

Due to the variability of solar power sources and increasing penetration of solar generation on energy grids, independent system operators and balancing authorities are facing a high level of uncertainty in an expected solar resource for managing the grid. Solar irradiance variability could be determined by both deterministic and stochastic signals. The deterministic signals have both seasonal and diurnal variation and can be determined using simple astronomical relationships. However, atmospheric conditions, such as water vapor, turbidity, and clouds are the most influential on the solar energy reaching the ground and they are variable in nature. The overall shape of solar energy production can be easily predicted for most of the time if the weather is clear from cloud cover, but significant errors in the level and timing of solar energy production are introduced by the passing of clouds that cause ramps in energy production. Therefore, site specific solar resource variability need to be predicted to help minimize the risk associated with the deterministically estimated solar energy production.

Forecasts of solar energy can be used to address expected variability and uncertainty in the solar resources and it is playing a key role in solar PV operation and management, accurate solar power dispatchability as well as scheduling. Independent System Operators (ISO) use day-ahead load forecasts to help schedule the amount of energy needed for each hour of the next day. Therefore, providing the ISOs and other stakeholders with an accurate solar energy forecast and associated forecast uncertainty will help to make better decisions about resource allocations to make an efficient integration of solar power in the energy market. Numerical Weather Prediction (NWP) are the most commonly used models to predict day ahead energy outputs from solar and wind energy. In any given weather forecast models, there are two factors that lead forecast skill to decrease as forecast lead-time increases: Uncertainty in the initial conditions and approximations in the numerical model development. Therefore, NWP based deterministic forecasts come with an error that cannot be avoided, but rather exponentially increase with increasing lead time. However, successful integration of solar power into electricity grids begins with a reliable day-ahead NWP forecast. Therefore, there is a need to provide predicted forecast uncertainties for better risk assessment for decisions based on forecasted energy values. In this work we have investigated the application of (NMV) model for a day ahead solar energy forecast uncertainty prediction.

II. dAta and methodology

For this study, measured and satellite based Irradiance data has been obtained for four NOAA SURFRAD stations: Desert Rock, NV, Fort Peck, MT, Goodwin Creek, MS, and Penn State, PA, for the period 1998 to 2015. The forecast models used in this study are derived from the European Center for Medium Range Weather Forecast (ECMWF), NOAA’s Global Forecast System (GFS) and National Digital Forecast Database (NDFD). The following equations are used to drive the NMV.

(3)

Where Kt is a clear sky index, KT* is an average clear sky index of a day (daily Kt) and

is called NMV. i is a time step. Ideally, the values of Kt vary from 0, when the weather is overcast to 1, when it is clear sky. The values of nominal variability ranges from 0 when it is cloudy or clear sky to very high values when the sky is partly cloudy. III. Results and discussion

Figure 1 shows the relationship between nominal variability and daily clear sky index (KT*) using ground and satellite based irradiance data for summer. The figure shows that the nominal variability is high when the KT* is between 0.4 and 0.7 due to scattered clouds. However, for overcast and clear sky conditions the variability is close to 0. A model obtained from this relationship is applied to a day head forecast to predict the uncertainty that is expected from the forecast model output, as shown in Figure 2 (relatively clear day) and Figure 3 (partly cloudy day). The results show that the variability model helps to predict the forecast uncertainty when used in conjunction with the day ahead deterministic forecast. This work also demonstrates the use of the NMV model for better risk assessment in the long term solar energy prediction by providing resource variability information at a site.