Pricing Algorithms in Short-term and Long-term Power Markets

Proces decentralizacije i dekarbonizacije elektroenergetskog sustava uzrokuje ekstremne volatilnosti na spot tržištu električne energije. Kako bi se sudionici na tržištu zaštitili tog od rizika i trgovali na učinkovito, važno je upotrijebiti prikladne matematičke modele za predikciju cijena kratkoročnih i dugoročnih tržišta. Glavne karakteristike cijena koje modeli moraju zadovoljiti su sezonalnost, mean reversion proces i iznenadne skokove u cijenama. U radu su obrađena tri modela, koja uspijevaju pokriti neke ili sve od njih. Prvi dva modela su jedno-faktorski modeli s Gaussovom razdiobom i sa inverznom Gaussovom razdiobom, bazirani na Ornstein-Uhlenbeck i Lévyevom procesu. Iako uspješno pokrivaju sezonalni trend i povratak u prosjek, ne uspijevaju replicirati skokove i osigurati odgovarajuću dinamiku cijena. Uvođenje dodatnog faktora u trećem modelu, baziranog na Poissonovom procesu, uspješno rješava problem iznenadnih skokova i povećava volatilnost u cijenama forward ugovora, no traži nešto kompleksniji način estimacije parametara.The processes of decentralization and decarbonisation of the electricity system causes extreme volatility in the spot electricity market. To protect market participants from price risks and make trading efficient, it is important to use appropriate mathematical models to predict the price on short-term and long-term markets. The main price characteristics that models must meet are seasonality, mean reversion and jumps process. The thesis deals with three models, which manage to cover some or all of them. The first two models are one-factor models with Gaussian distribution and with inverse Gaussian distribution, based on the Ornstein-Uhlenbeck and Lévy process. Although they successfully cover the seasonal trend and mean reversion process, they fail to replicate the jumps and ensure an appropriate dynamic price. By introducing an additional factor in the third model, based on the Poisson process, solving the problem of sudden jumps and increasing volatility in forward contract prices is done more successfully, with somewhat more complex way of estimating the parameters required