Algorithm to predict error in renewable energy forecasts.
In renewable energy, utility operators use forecast models to predict energy fluctuations over a future time horizon to account for operating reserves, and protect grid infrastructure from instabilities. These models are of limited accuracy and as a result operating reserves may be inadequate or over-provided, and grid instabilities may be caused by under or over production of power. Here we present a promising algorithm developed by a group of researchers led by Prof. Mahesh Bandi. The algorithm enables operators to manage varying levels of power generation and operate equipment efficiently and traders to make marketing decisions.
The technology is an algorithm that makes it possible to quantify two errors, used to qualify and improve forecast models, thus protecting against losses in revenue and grid instabilities caused by energy fluctuations. Specially, the algorithm utilizes generated and forecast time series for power generation derived from wind and analyzes temporal correlations in the wind fluctuations to quantify: (a) the forecast error defined by deviations between the high frequency components of the forecast and generated time series, and (b) a scaling error defined by a degree that temporal correlations fail to be predicted for an accurate predictor of wind fluctuations. Wind fluctuations may exhibit multi-fractal behavior at the turbine level and/or may be rectified to a fractal structure at the grid level. A memory kernel may be used to reduce the forecast and scaling errors.
- Wind Power
- Solar Power
- Ocean Power
- Minimal data is required to perform forecast error analysis; time series for actual power generated and forecast power
- Applicable to any data sampling rate