posted on 2024-11-12, 14:34authored byDissawe Mudiyanselage Lasanthika Harshani Dissawa
Power generation form solar photovoltaic has shown significant expansion over the past few years. The fluctuations in solar PV power output within short-term time frames are highly unpredictable and can create operational difficulties in local electricity grids. Particularly, the shadows of clouds and storm fronts create significant fluctuations in solar power being fed into the grid. Therefore, methods used to forecast solar irradiance at a ground level, and thus estimating the short-term PV power generation, are necessary to ensure effective management of solar PV connected to the grid. Since sky images truly reflect associated implications on a solar PV system, especially in terms of available power generation, sky image-based forecasting models provide reliable power generation forecasts for short-time intervals. Therefore, the implementation of a sky image-based forecasting model, while overcoming weaknesses in the existing models, is proposed in this thesis. As a key component of the research project, an inexpensive sky camera system was developed to obtain high-quality sky images, and to demonstrate and evaluate a sky image-based forecasting system. An on-site forecasting model was implemented using the sky images captured from the sky camera setup. A new cloud segmentation algorithm was introduced using YCbCr colour properties of sky images to identify both white and grey clouds for determining the cloud thickness. According to the newly proposed identification method, clouds were categorised into three categories and for each category, irradiance dropping factors were assigned. To find the individual cloud moving velocity, a new cloud motion tracking method was developed based on normalised crosscorrelation algorithm. From the devised cloud tracking method, the multi-layer cloud velocities were identified. Using the cloud motion vector details and cloud colour properties, the irradiance was forecasted for the location of the sky camera system. From this, the on-site forecasting model was further developed to obtain irradiance forecasts at multiple PV sites with the use of local cloud base heights obtained using a new approach based on stereographic method.
History
Year
2021
Thesis type
Doctoral thesis
Faculty/School
School of Electrical, Computer and Telecommunications Engineering
Language
English
Disclaimer
Unless otherwise indicated, the views expressed in this thesis are those of the author and do not necessarily represent the views of the University of Wollongong.