An Adaptive Dynamic Pricing Mechanism to Balance Supply and Demand in Electricity Markets Powered by Renewable Energy Sources

  • Mehmet Ali Ergun

Student thesis: Master's Thesis


According to the 2008 WWF living planet report, the United Arab Emirates has the highest per capita ecological footprint and a big portion of this ecological footprint is due to high electricity consumption. One way to reduce ecological per capita impact is to produce electricity by using renewable energy sources. However, intermittent nature of renewable energy resources makes it necessary for power grids to implement a different pricing strategy that can potentially alter the demand of electricity according to the available renewable generation capacity. In order to implement such a strategy, the electricity distribution company must determine the unit price of electricity in each time period. We focus on using learning agents that determine how this price should be set so that the total consumption of the consumers is aligned with the available renewable electricity production capacity. In this work, we define a structure of the electricity market in which the pricing mechanism consists of a distributor agent that facilitates dynamic pricing and an independent system operator (ISO) who oversees the distributor agents actions. A Q-learning distributor agent is tested with both consumer agents that have pre-defined behaviors and learning consumer agents. We define five metrics to assess the efficiency of the pricing mechanisms and evaluate the performance of our Q-learner according to these metrics. Our results show that the learning distributor agent paired with learning consumer agents lead to the most efficient market state compared to other cases we investigated. We also claim that learning distributor agents can also be used for predicting the amount of electricity needed for the next time period within an error band. Finally, we learned that, in mixed populations of automata and learning consumer agents, increasing the proportion of learning consumers in the system enhances other agents performances.
Date of Award2011
Original languageAmerican English
SupervisorJacob Crandall (Supervisor)


  • Electrical Engineering
  • Renewable Energy Sources

Cite this