Choose an application

• Reference Manager
• EndNote
• RefWorks (Direct export to RefWorks)

Mil Colinas is a NGO founded in 2011 in order to contribute to the battle against inequality. Mil Colinas mediates between northern and southern countries in order to further social development. It is an educative project which started with 30 children and has grown to almost 200 people in addition to three Rwandese educators. Due to the fact that the current room is too small to carry out the activities, there is a clear necessity for the association to have its own space.
Last summer (2015), I went to Rukara, the Rwandese village in which Mil Colinas operates. I began taking its measurement and shortly after the land was purchased. The current project plan is to utilize solar energy for the school's main source of power.
During the period in which I was there, several facilities were visited (mainly schools and the health centre of Rukara) and some of them are already working with photovoltaic energy. Meetings took place around the country with key energy businesses in East Africa, such as Mobisol. Also the educators and children who are involved in Mil Colinas were directly contacted in order to understand and record the needs and desires they have for the new school.
With all the necessary data from this African country, an off-micro grid has to be designed for the current project. This grid provides the required energy (mainly for lights and laptops) through a photovoltaic system and accumulators.
Initially, it was necessary to measure the consumption as well as the solar irradiation in the area. Several hypothesis were created; taking into account different weather and other assumptions.
Following the hypothesis, some simulations were run in Python in order to plot the charge level of the batteries per hour throughout the whole year. Having these results, some decisions could be taken in account in relation to the quantity of the photovoltaic panels and the batteries with the help of the tool Solver in Excel. The wiring was also calculated as well as the optimum angle for the electrical generators. The system is optimized regarding the economical plan. Several plans were made with AutoCAD are attached as well as the 3D views of the school made with SweetHome3D.
As result of this project, there is an installation which will work in 48 V before the inverter and in 230 V upstream; 10 photovoltaic panels of 295 W each and 12 batteries of 200 Ah of capacity. Each unit is part of the installation along with the inverter and the controller. Initial investment reaches $12,257.67 but it is necessary to have a budget of $19,800.89 in total to assure the proper function during at least 24 years.

Microgrid --- Off-grid --- Rwanda --- Rukara --- Africa --- Mil Colinas --- International Cooperation --- Education --- Photovoltaic energy --- Renewable energy --- Python --- Excel --- AutoCAD --- Sweet Home 3D --- Solver --- Business cases --- Afrique --- Coopération Internationale --- Énergie Photovoltaique --- Énergie Renouvelable --- Ruanda --- Energía Fotovoltaica --- Energía Renovable --- Casos de negocio --- Ingénierie, informatique & technologie > Energie --- Ingénierie, informatique & technologie > Ingénierie électrique & électronique

Choose an application

• Reference Manager
• EndNote
• RefWorks (Direct export to RefWorks)

The objective of this master thesis is to find transfer learning
methods which can be applied in reinforcement learning problems.
This method has to improves the learning speed of an agent on a
target task, given a set of already trained policies on similar
simpler tasks. We used Q learning algorithm combined to neural
network to represent Q function that determine how an agent act
on it’s assigned task.
Using source network trained on simpler problems, we manage to
find an architecture that combines networks with an interface to
transfer knowledge in the new network. This technique allowed us
to solve the OpenAI Gym mountain car problem using source
networks trained on a smaller mountain and a faster car.
We tested more complex connection architectures to improve the
representation complexity of the combined networks.
We also implemented a method to learn multiple networks in
parallel to exploit the ensemble properties when learning a new
task. This method improves the convergence and results.
Finally we tested our implementation on the OpenAI Gym
Cartpole environment with succes.

transfer --- learning --- mountain --- car --- cartpole --- neural --- network --- concatenation --- merge --- openai --- gym --- reinforcement --- Ingénierie, informatique & technologie > Sciences informatiques