Impact of uncertainties in meteorological databases on the design of off-grid microgrids and a new approach to ensure controlled sizing

Thanks to ongoing advances in renewable electricity technologies, microgrids offer a promising solution for supplying clean electricity to remote communities. In this context, equipment sizing is of crucial importance. To address this issue, designers minimize the cost of the electrical system by relying on historical weather records. This meteorological data is often sourced from satellite databases, as it is available for most locations around the world and covers a significant period of time. The present study highlights the discrepancies observed when using satellite data from different providers. Solving a microgrid optimal design problem as a benchmark, the use of different databases results in very significant cost variations (ranging from 65 to 143 k€) and, for certain design scenarios, may lead to situations where no energy is supplied. This finding leads to the proposal of five methods for merging satellite-derived data which can drastically reduce uncertainty across the solution space, when considered together in a decision support point of view. Based on a comparison with high quality data from a weather station, the proposed methods drastically reduce the range of potential optimal solutions, regardless of the climate under study. By switching from unmodified satellite databases to the suggested merged datasets, the average distance between the boundary of the feasible space and the reference is reduced on average by a factor of 3.2. More importantly, the standard deviation of the distance between the boundaries is reduced by a factor of 8.8 on average, which significantly reduces the variability and uncertainty of the input data. This finding is the key point of the paper.