Retailers and major consumers of electricity generally purchase a critical percentage of their estimated electricity needs years ahead on the forward markets. This long-term electricity procurement task consists of determining when to buy electricity so that the resulting energy cost is minimised, and the forecast consumption is covered. This decision-making problem is particularly challenging because of its sequential and highly stochastic nature, coupled with a poorly observable and potentially adversarial environment. Nowadays, the long-term electricity procurement task is generally performed by experienced consultants, based on customised rules and their expectations regarding the future energy market direction. The present scientific article proposes an alternative algorithmic solution which may interest these consultants, but also retailers and major consumers of electricity.
This research paper presents a novel algorithm, named Uniformity-based Procurement of Electricity (UPE), providing recommendations to either buy electricity now or to wait for a future opportunity, based on the history of calendar (CAL) prices. This algorithm is based on the idea that the purchase decisions should be split over the procurement horizon to spread the trading risk, with a nominal anticipation or delay depending on the market direction. The algorithm relies on a forecasting mechanism to predict the market trend and on the concept of procurement uniformity. This procurement uniformity level quantifies the deviation from a perfectly uniform reference procurement strategy purchasing a tiny amount of energy at each time step over the entire procurement horizon. Two variants were developed depending on the forecasting model considered, respectively UPE-MA for moving averages and UPE-DL for deep learning.
On average, both variants surpass the benchmark procurement strategies considered, and the top-performing UPE-DL algorithm achieves a reduction in costs of 1.65% with respect to a perfectly uniform policy achieving the mean electricity price. This represents an average yearly saving of €75,100 for an annual consumption of 100 GWh of electricity between 2012 and 2019. Moreover, in addition to automating the long-term electricity procurement task, the UPE-DL algorithm exhibits key advantages, such as a superior stability, with more consistent results obtained throughout the years. Another strength of this algorithm is the interpretability of the decisions advised, which improves the reliability of the procurement strategy and better facilitates its monitoring by a human supervisor. Finally, the approach depicted in this scientific article presents the major advantage of being easily adaptable to solve other commodity procurement problems presenting similar constraints.
Leave A Comment