AgentUDE17: A Genetic Algorithm to Optimize the Parameters of an Electricity Tariff in a Smart Grid Environment

Electricity retailers are the most vulnerable actors in the electricity grid since they are responsible for many financial challenges. As a business entity, electricity retailers aim to maximize their sales volume and minimize the procurement cost in a highly competitive environment. On the retail market side, they publish rich tariffs specifications, which resolves the needs of customers in time, energy, and financial domains. In the paper, we present an online genetic algorithm that optimizes the parameters of an electricity consumption tariff, such as unit retail price, periodic, sign-up, and early withdrawal penalty payments on the fly. Additionally, we use time-of-use (TOU) price scheme to reduce peak-demand charges. The algorithm was first deployed and tested in our winning broker agent (AgentUDE17), which competed in the Power Trading Agent Competition (Power TAC) 2017 Final games. In the paper, we first present the theoretical background and detail the concepts of the algorithm. Secondly, we comparatively analyze the tournament data. Post tournament analysis shows that AgentUDE17 successfully optimized its tariff parameters on the fly and significantly increased its utility. Additionally, it managed to reduce its peak-demand charges thanks to its adaptive TOU price schemes.