Trading strategies modeling in Colombian power market using artificial intelligence techniques

The aim of this paper is to present a model based on fuzzy logic and machine learning in order to maximize the profits of Colombian energy trade agents according to their risk profile. The model has two parts, the first one is a fuzzy expert system that gives a recommendation about the trade strategy these agents should follow, and whose definition depends mainly on market conditions. The second one is a reinforced learning mechanism with which the agents "learn" when they perceive the consequences of their actions, so they modify such actions looking for a reward not just in short but also in long-term. The whole model is validated using actual data as well as a simulation approach using synthetic time series for some relevant variables as hydraulic availability, energy pool price and bilateral contracts price.