The joint tasks 16 “Solar resource for high penetration and large scale applications” and V “Solar Resource Assessment and Forecasting” of the IEA Technological Collaboration Programmes respectively PVPS and SolarPACES started in July 2017 and will last for three years. The main goal of these joint tasks is to lower planning and investment costs for photovoltaic (PV) and concentrated solar thermal electricity (CSTE) systems by enhancing the quality the solar resource assessments and forecasting. Notably, one the subtasks is dedicated to enhanced solar datasets and bankable products: it aims at establishing methods to provide end-users with products and solar datasets along with their uncertainty from different sources of earth observation (ground measurements, satellite, numerical weather models) and combinations of them.
In particular, within the framework of this subtask, activities are conducted on automatic and expert-based quality control and gap filling methods dedicated to pyranometric and meteorological data used for solar resource assessment and forecasting.
The purpose of the workshop is first to propose a review of automatic quality control procedures (QCPs), including the emerging ones based on cross-comparison with surrounding in-situ measurements, models of clear sky irradiance and satellite based estimations (Urraca et al., 2017). Best practices and specific data visualization for expert based QCPs will be then gone over, in relation with typical measurement troubles (soiling, misalignment, time shift, etc.).
The second part of the workshop will be dedicated to gap filling methods to cope with data gaps in pyranometric and meteorological time series (missing data or detected as incorrect by the QCP). These gap filling methods are meant to complete the time series at the native time resolution or to aggregate at coarser resolution: e.g. averages from 1-min to hourly, from hourly to daily, etc. First, a review of such gap filling methods, including some using satellite-based estimations will be exposed. Second, a Monte-Carlo based benchmarking methodology based on reference time series with randomly drawn missing data will be described. This benchmark will be applied to a selection of gap filling methods and relative performance will be discussed, notably owing to temporal resolutions, classes of temporal variability and percentage of missing data.