EPSO in energy retail market simulation on an intelligent agent platform (JADE)

Agents representing distribution companies must optimize their market strategies against competitors. In each move, an agent simulates a number of periods in the future and selects the next move in order to optimize the forecasted results. Agents with distinct optimization algorithms competed against each other and EPSO emerged as the winning algorithm.

The figure shows the weekly profits of agents equipped with EPSO, PSO, SSGA – Steady State Genetic Algorithm, DCGA – Genetic Algorithm with Deterministic Crowding and MPGA – Multi-Population Genetic Algorithm. The simulation run for a period of 2 years with a weekly re-optimization of the marketing strategy extending for a period of 2 months. The competition was made fair by balancing for each algorithm the computing effort in terms of calls to the fitness function (maximizing the retailer profit).