Week 15: Perfecting the algorithms

In this week we had a meeting with our inspirator, Professor Zita Vale to present the work developed until the moment and make some questions regarding the methods that were being implemented in the PSO for the community energy management. The meeting went well and all the suggestions received have been implemented in the project.

Still regarding this question, the team is now taking two different approaches when managing the community energy: The first where the optimization is done for one period only, by using data of whole day and the second where one day is optimized but a day and a half is used as an input.

Meanwhile, regarding the generation and consumption predictions, in this week, several strategies were used to try to improve the consumption models. Columns for classifying the day of the week were created, as well as context column, which are features that divide the day in several periods where people might be conducting a specific activity such as breakfast or lunch, and that help classify those types of periods.

It was also noticeable that some of the predictions were suffering from lags and to avoid that, less importance was given to the features that represented old consumptions. The results improved, though not by a lot.

Regarding the integration of both modules, the integration code has been developed and a UI is being created to help demonstrate the functionalities of the project.