Using Large-Scale Circulation Index History to Put Energy Demand and Renewable Supply Forecasts in Context

Energy weather forecasters are strongly motivated to wrest predictability from any available weather data that exhibits some degree of memory.   Here we present a set of tools that allows forecasters to place long-range forecasts of large-scale circulation indices, as well as forecasts of sensible energy-related weather such as regionally aggregated heating or cooling degree days and solar and wind generation, in the context of the past evolution of the predicted index value.  Past analogs are chosen based on the similarity of a set of user-chosen circulation indices (ENSO, MJO, AO, PNA, etc.).   The analog-generation algorithm performs statistical tests that determine the most appropriate timescale for comparison (dependent on the decorrelation timescale of the index in question) and the appropriate weighting of the indices (based on their relationship to the forecast variable).   

The upper panel of the accompanying figure shows an example for the AO index, setting a forecast of the AO index by the NOAA GFS forecast model in the context of the 5 periods most similar in evolution to the present (including the initial forecast day) in a 65-year record of daily values.  We present analysis of any predictive skill gained from the analog ensemble relative to the dynamical forecast alone, and explore applications to select analog model forecast ensembles.  The lower panel shows the climatological dependence of wind generation in each RTO on the AO index, derived from a simulation using 37 years of the Climate Forecast System Renalysis.   The bar charts show wind generation for 11 bins of daily AO values and show a tendency for more wind at the high and low end of the AO distribution, less in the middle range.  Analysis of such dependencies can be used to gain insight into  the co-varibility of various components of our energy system with the various leading modes of atmosheric variability.