STTR Phase I: Solar Irradiance Microforecasting

The broader impact/commercial potential of this project to develop short term Solar irradiance forecasting, will be to support very large deployment of Solar photovoltaic (SPV) generation capacity, by reducing the cost of mitigating cloud caused fluctuation of SPV electricity generation. This increased SPV system deployment will reduce the amount of base load and peaking generation from greenhouse gas causing, and water consuming fossil fuel generators. Such forecasting will enable development of pre-­‐ mitigation strategies instead of post mitigation using electrical storage systems. Prior studies indicate that this will result in the reduction by up to a factor of five, of the input/output requirements of the electrical storage system used in the pre-­‐mitigation scenario, compared to the post mitigation scenario. These benefits will be seen with grid-­‐tied, micro-­‐grid and off-­‐grid SPV systems. This opens commercial opportunities for introducing intelligent sensors and control systems to reduce bulk electrical storage. The technology areas used in this project include sensors, 3D printing, neural network based learning systems, embedded computers and cloud computing. The market sectors that will see a positive impact include all demographics as consumers, and manufacturers of SPV modules and SPV balance of system suppliers. This Small Business Technology Transfer (STTR) Phase I project addresses the problem of mitigating cloud movement induced fluctuation in the output of SPV systems. The research objectives of Phase I are (a) prototype a whole sky imager that provides sufficient circumsolar image discrimination, to drive a neural network based learning system ? this will require development of a 3D-­‐printed mounting system for a whole sky sensor, and interface to a cloud connected, local single board computer, (b) develop and optimize Image Acquisition, Compositing, Analysis, and Forecasting Algorithms to provide 15-­‐500 second forecasts of Solar irradiance, and (c) deploy imager + software prototypes to evaluate real live sky imagery in multiple locations with different weather patterns, by gathering data to ?train? the neural network. It is anticipated that this evaluation and analysis of prototype performance will continue in subsequent phases, to obtain high confidence results. The anticipated results of the research in Phase I are (i) refinement of the image capture system to produce ?good? imagery, (ii) development of procedures to tune neural network learning system towards obtaining high confidence forecasts, and (iii) understanding of performance requirements of local single board computer.