A Novel Semiparametric Structural Model for Electricity Forward Curves

Abstract

We propose a novel semi-parametric structural model to estimate the electricity forward curves based on elementary forward prices. The proposed model (i) explores the nonarbitrage relations between contracts with overlapping delivery periods, (ii) considers a parametric structure for price seasonality and exogenous variables, and (iii) uses non-parametric techniques to extract the remaining inter-temporal and cross-maturity information from data. Thus, our model allows users to estimate and complete the historical prices of any swap contract. We address the multi-objective estimation problem by hierarchical optimization. First, arbitrage levels are minimized. Then, the parametric part of the model is estimated. Finally, smoothness in the maturity and trading date dimensions are jointly considered in the estimation of the non-parametric part of the model. Based on a controlled study with real data from the Nordic power market, we show that our model outperforms benchmarks in terms of estimation error for missing data. We also isolate the effect of accounting for overlaps and smoothing in the trading dates dimension. Results show that these two key features of our model are crucial for improving the model accuracy. Finally, we apply our method to estimate the Brazilian forward curve and reconstruct the historical data.