This research focuses on demand side management in Smart Grids and the hypothesis of reducing peak demand using Smart Grid capabilities.
Alongside with the production of electricity, concerns related with the efficiency of production, distribution and consumption of produced energy appeared. These concerns arise from the willingness of producers to maximize profit and environmental awareness, which is growing every day in our society.
Driven by that motivation, research in renewable energy resources is increasingly augmenting and potentiating the appearance of new challenges in the production of these cleaner energies, that in addition to be greener are also cheaper in a long-term. One of the main challenges is powering all demand with these energies. Renewable energy generators have a long setup time and it proves to be difficult since in peak situations, electricity delivery must be instantaneous, making them dependent on faster delivery time petrol generators to manage peak demands. Managing demand peaks require control of consumer devices which can only be possible nowadays using Smart Grid capabilities in order to communicate with consuming devices. This approach also demands a certain flexibility of users to postpone or anticipate appliance executions, having as counterpart cheaper energy prices in certain times of the day.
In this research is assumed that electricity prices are known 24 hours in advance, making it possible to schedule home appliances operation. Therefore, using communication abilities of a Smart Grid and electricity prices, this research sets as a main goal to develop an algorithm that can schedule devices in order to help reduce peak demand. This scheduling is constrained by user input, indicating the time frame within which each schedulable device must execute.
The resulting scheduling algorithm is based on a meta-heuristic called Evolutionary Algorithms, which uses as a solving technique as a metaphor of human evolution, by trying to mimic crossover between individuals and possible mutations that also happened during human evolution. This method allows finding very good solutions within a reasonable amount of time, making it feasible for a real-world operation. Results are obtained within milliseconds, which for human perception is almost instantaneous.
All goals proposed in this master thesis were successfully completed. Results are promising in terms of employing the proposed algorithm in the production phase of its parent project.
For any additional information regarding this project, please contact us using the inquiries form.