Performance assessment of intra-day solar irradiation forecast in Uruguay using satellite cloud motion vectors

Abstract

Reliable solar irradiation forecast is increasingly important for solar PV power dispatch. Satellite forecasting methods are suitable for the intra-day time horizons. In this work we evaluate the performance of the satellite cloud motion vector (CMV) forecasting technique proposed by Lorenz et al. (2004) for the Uruguayan territory. The local implementation of this technique is also presented in this article. The performance assessment is done for the prediction of the hourly solar global irradiation at ground level from 1 to 5 hours ahead and for site and regional forecast. The relative root mean square deviation (rRMSD) ranges from 19% to 37% for the site CMV forecast and from 9% to 18% for the regional CMV forecast. As expected, the deviations in the prediction are higher over specific sites than over the entire region. Positive forecasting skills (FS) are obtained for all the inspected time horizons in both spatial scales (site and regional). The FS metric peaks at +16% for site forecast (at 2 hours ahead) and peaks at +30% for regional forecast (at 4 hours ahead).