Operational Shortest-Term PV Solar Forecasting for ramp rate control with an ultracapacitor energy storage system using a Whole Sky Imager

Abstract:

UCSD has partnered with Maxwell Technologies to demonstrate Maxwells’ ultracapacitor energy storage system (UESS) using UCSDs’ shortest-term advective forecast for PV systems. Specifically, UCSD will be supplying 5-minute forecasts to predict ramp events for the UESS, which will then discharge/charge the system as appropriate for the event. Four different metrics will be used to evaluate the effectiveness of the UCSD advective forecast with the UESS: (1) The root mean square error, root mean bias, and root mean absolute error will be calculated for the 5-minute forecast using measured irradiance from the UCSD DEMROES stations and compared to a persistence forecast (2) A “matching” error analysis will be performed to compare the 5-minute forecasted cloud cover of the PV system to the actual cloud cover at the forecasted time (3) The matching error of the advective forecast will be compared to the matching error of a persistence forecast to determine if, operationally, advective or persistence forecast performs best (4) Timing of predicted ramp events using the advective forecast will be compared to actual ramp events experienced by the UESS.

The above metrics will also be used to analyze the effectiveness of cross-correlational and optical flow advective schemes in an operational setting. The cross-correlational method analyzes images from two different times to find an average velocity vector for cloud cover. Optical flow uses images from two time steps to find a velocity vector for each pixel of an image, allowing different sections of clouds to move at different speeds and directions. Hence, it is hypothesized the optical flow advective scheme will perform better then the cross-correlation method in operational settings.