Modelling of an electromagnetic generator using artificial neural networks

• Main Author: Rocha, Renato Miguel Emílio
• Publication Date: 2018
• Format: Master thesis
• Language: eng
• Source: Repositórios Científicos de Acesso Aberto de Portugal (RCAAP)
• Download full: http://hdl.handle.net/10773/23608
• topic: Redes neuronais; Eficiência energética; Arquiteturas de redes de computadores

The increasing necessity for autonomous functioning systems alied with comsumption reduction by microelectronic devices over the last years, has motivated the research on self-powering devices for remote applications. Developing an energy harvestring device for a determined application requires its study and optimization. Therefore, modeling the dynamics of the system for simulation purposes becomes mandatory. The use of analytical mathematical models or FEM is a standard approach for the development of computable models. However, this approach reveals to be time-consuming due to temporal restrictions established not only by the model development but also by its simulation. This work investigates the application of Artificial Neural Networks on the modeling of magnetic levitation systems for energy harvesting purposes. The data collection implied by a Neural Network approach demanded the development of a device suitable for the acquirement of information intrinsic to the system. A testing station was built with the goal to induce rotational excitations on the device and acquire the motion dynamics of the levitation magnet. Different network architectures and implementation techniques are approached in this work in order to optimize the characteristics of the model. From the different approaches taken for proper model implementation, the configuration named in this work as NARX BROC allowed the attainment of correlation values above 95% for simulations inside and outside the training range, when compared with experimental results. Also, the performance of the developped generator is analyzed and discussed according to intended applications