Design and Evaluation of the Distributed Solar Power Production Forecast Component of the Solar and Wind Integrated Forecast Tool (SWIFT)

- Indicates paper has been withdrawn from meeting
- Indicates an Award Winner
Tuesday, 6 January 2015: 2:00 PM
224B (Phoenix Convention Center - West and North Buildings)
John Zack, AWS Truepower LLC, Albany, NY; and K. Waight, S. Young, and E. J. Natenberg

The management of solar and wind power variability on small island grid systems is an extremely demanding task due to the small size of the systems and the lack of interconnections. The Hawaiian Electric Company (HECO) family of utilities operates five island grid systems with a high and growing penetration of solar and wind generation. In order to provide a tool to assist in the management of the variability, a Solar and Wind Integrated Forecast Tool (SWIFT) has been developed to provide guidance for the management of solar and wind power variability. SWIFT is built on the multi-method ensemble approach to forecasting in which the prediction that is ultimately presented to the application is based on an optimized composite of the individual methods. One of the most critical and challenging applications of SWIFT is the prediction of the electric power produced by distributed photovoltaic (DPV) resources. This application is becoming critical for HECO because of the very high level of penetration (generally more than 20% of midday demand) on the island grid systems. For example, on the island of Molokai the midday load is typically in the 3 to 4 MW range and the DPV capacity is over 2 MW. The other side of the problem is the set of significant challenges that must be overcome to predict the actual DPV generation. In addition to the meteorological challenges of forecasting the small-scale variations in cloud coverage and attributes, there is a scarcity of information about the DPV attributes (panel type, orientation etc.) and the actual generation from the systems. Thus, methods must be devised to infer the generally unmeasured DPV generation from the measured or estimated solar irradiance. SWIFT generates 0-6 hour and 0-48 hour deterministic and probabilistic forecasts of solar and wind power production. The DPV production forecasts are generated for three levels of aggregation: (1) individual substation (i.e. all of the DPV on the circuits attached to that substation), (2) geographical regions defined by clusters of substations, and (3) system (island)-wide, which is the collection of all of the regions on a specific island. Individual DPV power output models have been formulated for each substation based on the available measured generation and meteorological data for selected points and DPV systems on each island. The presentation will provide (1) an overview of the solar and wind penetration resources on the HECO grid systems and the associated operational issues that have been encountered; (2) a description of the DPV prediction component of SWIFT; (3) initial results from the statistical evaluation of the DPV production estimates and forecasts; (4) examples of the use of DPV forecasts in critical situations faced by grid operators; and (5) a preview of the plans for the integration of the forecast information into the EMS as part of the DREAMS project under support by the US Department of Energy.