Banca de DEFESA: Matheus José de Carvalho

Uma banca de DEFESA de MESTRADO foi cadastrada pelo programa.
STUDENT : Matheus José de Carvalho
DATE: 12/07/2023
TIME: 14:30
LOCAL: Plataforma Microsoft Teams
TITLE:

Methodology for Object-Position Determination in Space from Multiple-Camera Images for Use in Short-term Solar Forecasting

KEY WORDS:

Cloud height; Stereo Vision; Fisheye Camera; Intra-Hour Solar Forecasting

PAGES: 89
BIG AREA: Engenharias
AREA: Engenharia Mecânica
SUMMARY:

Cloud dynamics is relevant for several areas, such as meteorology, agriculture, and airport operation, among others. In particular, for short-term solar forecasting, important for the management of solar power plants and distribution networks with high participation of solar energy, a more accurate characterization of cloud motion can improve generation forecasting as their presence is the major responsible for the variability of irradiance on intra-day time scales. Thus, in this paper, a methodology is proposed that uses triangulation of images from two fisheye-lens cameras to measure the position of objects in space, aiming at its use in short term solar forecast platforms based on sky images. A two-camera system, together with a wireless router, was installed on the roof of the Energy and Environment Laboratory of the University of Brasilia. The methodology developed is based on a stereo vision model, used to process the images and determine the location of visible objects in the images. The proposed model for image processing does not require the removal of image distortion for pinhole images, normally used in fisheye cameras, nor the rectification of the original images, being these the main contributions compared to other existing models. For the validation of the model, a drone was used as an object with known position. For this, the drone, in stationary night flight, was positioned at various points and altitudes desired and, with its lower light on, images of the sky were captured with the cameras. The coordinates of the drone in the images obtained by the model were then verified and compared with those reported by the drone. The results showed an average percentage error around 10% (relative to the value reported by the drone) at altitudes ranging from 90 to 490 meters. These values are acceptable for the desired application, with measurement uncertainties comparable to cloud height obtained by ceilometers, with the advantage of enabling a wide field of view, and with lower computational costs than traditional models.

COMMITTEE MEMBERS:
Externo à Instituição - ENES GONÇALVES MARRA - UFG
Interno - 3138349 - EDGAR AMARAL SILVEIRA
Presidente - 2245668 - MARIO BENJAMIM BAPTISTA DE SIQUEIRA
Externo ao Programa - 3295206 - RAFAEL CASTILHO FARIA MENDES - null