Steadysun at the 7th International Conference Energy and Meteorology

Like many islands, Tahiti (French Polynesia), is heavily dependent on hydrocarbon imports. To decarbonize the electricity generation and reduce the Polynesians' energy bill exposure to the volatility of global hydrocarbon prices, the country’s energy transition strategy strives to increase the share of renewable energy (RE) – encompassing hydroelectricity and solar power – in its production mix from 37% in 2020 to 75% in 2030. However, due to its variability, integrating solar photovoltaic (PV) into a non-interconnected grid, poses significant challenges in maintaining power system stability while optimizing the profitability of solar assets and storage systems, from development to operation.

During the poster sessions, Guillaume presented an overview of our array of services in the region. Over the past six years, we have been actively collaborating with the local grid operator (EDT Engie) as well as with several developers of hybrid power plants (PV+storage projects) to facilitate the integration of solar PV. Our approach leverages Physics-AI modeling, utilizing diverse input data such as global and regional numerical weather prediction model (NWP) outputs, satellite imagery, data from a network of 8 all-sky imagers across the island, and production measurements from various plants. The result is a tailored solution furnishing 12 hours and 10 minutes ahead forecast data for total solar PV generation. This solution bolsters daily energy dispatch operations and spinning reserve management. Furthermore, our efforts extend to optimizing project profitability at various stages of development. We have conducted thorough studies, evaluating the spatio-temporal variability of solar resources on the island, sizing storage systems according to local grid codes, and accurately assessing the energy yield of PV+storage systems using historical and forecast data.

Subsequently, during the parallel session on Energy Operations and Maintenance, Damien gave an insightful presentation on Day 2. He delved into our cutting-edge intraday forecasting solution, and explained its current capabilities and limitations. As part of the MoNuTeR project , a dedicated 1km-WRF model has been carefully configured and calibrated to simulate cloud cover and solar radiation in the best possible way, both of which are strongly influenced by intricate local atmospheric phenomena (e.g. convective systems). The blending of this dedicated simulation with other classical NWP outputs, leads to a high performance forecasting solution for Tahiti. The confidence in the forecast is quantified through a P20-P80 uncertainty interval as well as a custom binary confidence indicator, which draws attention to the risk of significantly higher/lower PV production in specific timeframes. This indicator captures uncertainty stemming from the location/timing of convective systems by comparing local predictions with predicted meteorological conditions around the island. A live correction mechanism, leveraging real time PV measurements, further refines the forecast up to 4 hours ahead.