Electricity Price Curve Modeling by Manifold Learning

This paper proposes a novel non-parametric approach for the analysis and prediction of electricity price curves by applying the manifold learning methodology. Cluster analysis based on the embedded low-dimensional manifold of the original price data is employed to identify characteristics of the price curve shape. The proposed price curve model performs well in forecasting both short-term price such as the day-ahead prices and longer term price such as the week-ahead prices. The forecast accuracy is demonstrated by numerical results using historical price data taken from the Eastern U.S. electric power markets.