The accelerated installation of solar power systems especially distributed photovoltaic systems (PV) has imposed severe operational challenges on local and regional power grids due to the intermittent and uncertain nature of ground-level solar irradiance. High-fidelity directional solar irradiance (or plane-of-array POA) forecasts are enabling and cost-effective technologies to mitigate the negative effects on the power grids caused by PV power variability. In recent years, solar irradiance forecasting methodologies for intra-hour, intra-day and day-ahead energy markets have been extensively explored using data-driven, local-sensing, remote-sensing, numerical weather prediction and hybrid methods. However, the accuracy of forecasts has begun to asymptote and improvements have tapered recently due to the insufficiency of relevant and high-resolution data, and simplified prediction of cloud dynamics due to its rapidly evolving characteristics in the complex atmospheric system.
Therefore, this project proposes to collect and utilize new data, and develop new model in the aim to improve cloud cover forecasts and intra-hour POA forecasts. The focal objective is to improve the intra-hour POA forecast skill from 20% to 30% and improve ramp forecast accuracy from 50% to 70% during high variability periods. To achieve the objective, cost-effective solar instruments and programing package will be developed to automatically collect high resolution local-sensing data and remote-sensing data, for the purpose of atmosphere and clouds surveying. Secondly, a comprehensive hybrid model with the integration of atmospheric radiative transfer, local-sensing, remote-sensing and deep learning techniques will be developed to forecast cloud cover. The hybrid approach is expected to retrieve detailed cloud motion and optical properties, which will potentially improve the prediction of future cloud field. Thirdly, a full 3D Monte Carlo radiative transfer model will be developed to quantify the effects of finite clouds (with diverse locations, geometries and optically properties) on local POA with varied panel orientations. Then a thorough framework will be developed to forecast POA based on the predicted cloud cover and evaluate the forecasting model. In sum, this project proposes to utilize new data and develop new model to theoretically investigate the radiative interactions within the complex atmospheric system and to advance the research frontier of intra-hour forecasting of directional solar irradiance.
