Towards a standardized procedure to assess solar forecast accuracy: A new ramp and time alignment metric. (1st July 2017)

• Record Type: Journal Article
• Authors: Vallance, Loïc; Charbonnier, Bruno; Paul, Nicolas; Dubost, Stéphanie; Blanc, Philippe
• Is Part Of: Solar energy. Volume 150(2017)
• Page Start: 408
• Page End: 422
• Publication Date: 2017-07-01
• Subjects: Solar irradiance -- Forecast -- Metric -- Performance -- Classification; Solar energy -- Periodicals; Solar engines -- Periodicals; 621.47
• DOI: 10.1016/j.solener.2017.04.064
• Languages: English
• ISSNs: 0038-092X

Abstract

Highlights: Highlighting the limitations of the RMSE in concrete situations. Proposition of a new metric describing the temporal alignment of a forecast (TDM). Proposition of a ramp score quantifying the ability to forecast ramps. Suggestion of a general procedure to assess the quality of a forecasting method. Adaptation of a classification of instants to assess the categorical performances. Abstract: The temporal variability of the solar resource – which occurs at different time scales – is a major concern regarding its impact on the grid and photovoltaic (PV) power plants. To handle the increasing penetration rate of solar energy, grid managers require accurate forecasts of incident solar irradiance. This work discusses standard and new procedures to assess the quality of solar forecasting models and highlights the limitations of some widely used metrics in concrete situations. The paper recommends practices for characterizing the quality of a forecast, i.e. a quality control to reject any suspicious data; classifications to relate each performance criterion to the nature of the solar irradiance variability; and a wide selection of established and new metrics. In this context, two new metrics are notably proposed in order to obtain more complete information about the performance of a forecasting method. This article aims to make a step towards standardizing metrics for solar forecasting by taking into account some of the new metrics recently presented in the literature.