Analog Ensemble Methodology: Expansion and Optimization for Renewable Energy Applications

Abstract:

Renewable energy is fundamental for sustaining and developing society. Solar and wind energy are promising sources because of their decreased environmental impact relative to conventional energy sources, improved efficiency, and increased use. A key challenge with renewable energy production is the generation of accurate renewable energy forecasts at varying spatial and temporal scales to assist utility companies in effective energy management. Specifically, this research applies the Analog Ensemble (AnEn) methodology to short-term (0-48 hour) wind speed forecasting for power generation and short-term (0-72) hour solar power measured (PM) output predictions. AnEn uses a set of past observations corresponding to the best analogs of a deterministic numerical weather prediction model to generate a probability distribution of future atmospheric states: an ensemble of analogs. Currently the AnEn methodology equally weights predictors and only handles 1D(time). We determine an optimal distribution of predictor weights based upon parameter characteristics, investigate spatial variations in the application of the methodology and develop a theory expanding the methodology into 2D. The AnEn methodology improves short-term prediction accuracy, decreases computational costs and provides uncertainty quantification allowing utility companies to manage over- or under power generation for renewable energy sources.