Net interchange schedule forecasting using Bayesian Model Averaging

The future power grid will need to incorporate systems and processes with a higher degree of variability and randomness due to the penetration of renewable energy resources and the increase of energy demand. Forecasting variables in a more uncertain environment poses new challenges and revisions of the existing forecasting methodologies will have to be made to maintain forecasting accuracy. This paper investigates an ensemble-based technique called Bayesian Model Averaging (BMA) to improve the performance of Net Interchange Schedule (NIS) forecasts. BMA is used to combine an ensemble of five diverse forecasting methods that each estimate NIS. The results, which examine performance for two separate years of real-world NIS data, demonstrate that BMA's aggregated forecasts reduces forecasting error by 30-55% in comparison to all individual prediction methods. This work illustrates a new possible mechanism for improving NIS forecasting accuracy, as well as other power grid system variables, and lays the foundation for future work on aggregate models that can balance computational cost with prediction accuracy.