Expressing Solar Power Forecast Uncertainty

As we advance the MDA Solar Power Forecasting System, each year at this annual meeting we present new lessons learned. Our multi-model ensemble system has been capable of providing uncertainty information and finally we are starting to see more interest in this information from our forecast clients and in forecast trials as well as in funding opportunities oriented toward utilizing forecasts in grid integration. For solar farm and RTO aggregations, we have provided percentiles (equivalently, probability of exceedance) forecasts, probability bins (probability generation will be within a particular range), and the minimum and maximum 5-minute averages within an hour. We will show validations of these and compare and contrast the information content readily gleaned from these different approaches to expressing forecast uncertainty. Also, we have an interesting case study when a thunderstorm complex erupted over a cluster of solar farms, having a large impact on the RTO-wide totals due to poor geographic diversity of large utility-scale solar farms. The case study illustrates the comparison between individual models at various lead times in picking up this event and how that is reflected in the probabilistic forecast information.