The objective of this thesis will be to explore the applicability of deep neural networks as appliance detectors. The student will be provided with training data, which includes the energy consumption of typical household appliances over the time span of one year. By using the data, the student will first label the data, train the DNNs and evaluate their performance. Depending on the thesis type, the student will have to compare the performance to another appliance detector. The selected approach will be implemented in Python or C++.
Contact: Christoph.Klemenjak@aau.at
Type: Scalable to all thesis types
Supervised learning techniques for Energy Advisors
Energy Advisors such as Mjölnir provide valuable feedback to the user. The feedback builds on gathered knowledge and observations of the energy consumption in households. The objective of this thesis will be:
- Review applicable supervised machine learning techniques and discuss their application in Energy Advisor tools
- Implement the most preferable technique and embed it into the Mjölnir framework
- Evaluate the performance by means of a case study
Contact: Christoph.Klemenjak@aau.at
Type: Scalable to all thesis types
Forecast of energy consumption in the residential sector
The energy consumption of users can be seen as an aspect of human behaviour. Without a doubt, this behaviour is influenced by weather conditions. When trained with smart meter readings, neural networks can be applied to predict the energy consumption of households. The question is, if weather forecast can serve as adequate training data to successfully predict the energy consumption of households. The scope of this thesis will be to explore this question.
Type: Scalable to all thesis types
Energy advisor platform
Mjölnir is an open-source energy advisor platform, where different energy consumption feedback mechanisms can be implemented and assessed. The advisor platform analyses and processes energy consumption data, which was collected by measurement hardware such as smart meters or power meters.
The objective of this thesis will be to enhance the advisor platform (adding new widgets) and to explore new feedback mechanisms. The student(s) should be familiar with HTML, PHP, JavaScript or Python.
For more information about the advisor see: http://mjoelnir.sourceforge.net/ and https://arxiv.org/abs/1505.0131
Contact: Christoph.Klemenjak@aau.at
Type: Bachelor/RP/Master
Load Disaggregation: WebApplication
Nonintrusive Load Monitoring (NILM) identifies used appliances in a total power load according to their individual load characteristics. The aim of this thesis is to build up a website which can import the measured power consumption of users, disaggregate the used power profiles of appliances, share power profiles and calculate the power costs for each appliance. We are looking for a student which is fit with HTML5 and object-oriented programming to implement on the one hand the homepage and on the other hand the load disaggregation algorithm.
Further Literature
Contact: Christoph.Klemenjak@aau.at
Type: Bachelor/RP/Master
No comments:
Post a Comment