A carregar...
Publicação
Infraestrutura De Inteligência Artificial Para Eficiência Energética
| Nome: | Descrição: | Tamanho: | Formato: | |
|---|---|---|---|---|
| 4.12 MB | Adobe PDF |
Autores
Orientador(es)
Resumo(s)
A herdade São Lourenço do Barrocal, local onde foi desenvolvida esta dissertação,
apresenta grandes consumos de energia elétrica ao longo de todo o ano, uma vez que as suas
áreas de desenvolvimento são o turismo e a hotelaria.
Posto isto, a presente dissertação teve como objetivo principal a implementação de
uma infraestrutura que permitiu monitorizar os consumos elétricos em alguns espaços da her-
dade. Além deste objetivo e, recorrendo a algoritmos de
Machine Learning, foi ainda desen-
volvida uma análise sobre a contribuição de diversas variáveis, como por exemplo, dados at-
mosféricos, dados de ocupação da herdade, na previsão de consumos.
A abordagem utilizada para resolver o objetivo principal desta dissertação consistiu na
implementação de sensores Shelly nos vários quadros elétricos e na instalação de uma base
de dados
Time-Series (a InfluxDB), de modo a guardar os dados provenientes dos sensores,
dos dados atmosféricos e dos dados resultantes da atividade hoteleira. De modo a monitorizar
os dados obtidos, foi desenvolvido um
Dashboard através do
software Grafana no qual foram
criados vários gráficos com o objetivo de apresentar o desempenho da herdade.
Por fim, com base em todas estas informações, foram desenvolvidas, recorrendo a al-
goritmos de
Machine Learning (Multilayer Perceptron, Support Vector Machine, K-Nearest
Neighbors, Random Forest e Linear Regression), previsões dos consumos elétricos da herdade.
Com estas previsões foi possível avaliar individualmente cada algoritmo utilizado, sendo que
o algoritmo Random Forest foi aquele que apresentou melhores resultados ao longo destes
processos preditivos.
The São Lourenço do Barrocal estate, where this dissertation was carried out, consumes a lot of electricity throughout the year, since its main areas of development are tourism and the hotel industry. That said, the aim of this dissertation was to implement an infrastructure that made it possible to monitor electrical consumption in certain areas of the estate. In addition to this objective, using Machine Learning algorithms, an analysis was also carried out on the contri- bution of various variables, such as atmospheric data and occupancy data on the estate, in predicting consumption. The approach used to solve this problem consisted of implementing Shelly sensors in the various electrical panels and installing a Time-Series database (InfluxDB) to store the data stemming from the sensors, from the atmospheric data and from the hotel activity data. To monitor the data obtained, a Dashboard was developed using the Grafana software in which various graphics were created to show the behaviour of the estate. Finally, based on all this information, predictions of the estate's electricity consumption were developed using Machine Learning algorithms (Multilayer Perceptron, Support Vector Machine, K-Nearest Neighbours, Random Forest, and Linear Regression). With these predic- tions, it was possible to evaluate each algorithm individually, with the Random Forest algorithm showing the best results throughout these predictive processes.
The São Lourenço do Barrocal estate, where this dissertation was carried out, consumes a lot of electricity throughout the year, since its main areas of development are tourism and the hotel industry. That said, the aim of this dissertation was to implement an infrastructure that made it possible to monitor electrical consumption in certain areas of the estate. In addition to this objective, using Machine Learning algorithms, an analysis was also carried out on the contri- bution of various variables, such as atmospheric data and occupancy data on the estate, in predicting consumption. The approach used to solve this problem consisted of implementing Shelly sensors in the various electrical panels and installing a Time-Series database (InfluxDB) to store the data stemming from the sensors, from the atmospheric data and from the hotel activity data. To monitor the data obtained, a Dashboard was developed using the Grafana software in which various graphics were created to show the behaviour of the estate. Finally, based on all this information, predictions of the estate's electricity consumption were developed using Machine Learning algorithms (Multilayer Perceptron, Support Vector Machine, K-Nearest Neighbours, Random Forest, and Linear Regression). With these predic- tions, it was possible to evaluate each algorithm individually, with the Random Forest algorithm showing the best results throughout these predictive processes.
Descrição
Palavras-chave
Sustentabilidade Internet of Things Machine Learning Shelly Dashboard Grafana