A Class of Marked Point Processes for Modelling Electricity Prices

This paper presents a family of processes to model electricity spot prices in deregulated markets. Besides mean-reversion, a property they share with other comodities, power prices exhibit the unique feature of spikes in trajectories. We introduce a class of discontinuous processes exhibiting a « jump-reversion » component to properly represent these sharp upward moves shortly followed by drops of similar magnitude. Our approach allows to capture – for the first time to our knowledge – both the trajectorial and statistical properties of electricity pool prices. The quality of the fitting is illustrated on a database of major US power markets.