| Turkey is located between 36 and 42 N latitudes, meaning it is close to axis of the equator. Therefore, Turkey is very rich in terms of the potential of renewable energy resources, although an important part of this potential is not active yet. The share of renewable energy sources in the total production of electrical energy was 25.2% in 2011. The share of wind and PV are 2.1% and 1% in production electricity. The role of PV systems is in the infant stage. Energy gap between production and consumption has been increasing in the world. The reason for the lack of PV (Photovoltaic) and wind systems is financial feasibility. Renewable resources such as PV systems and wind power systems are built considering their efficiency of output power. Solar irradiation is one of the major renewable energy sources, but the forecasting of solar irradiation depends on meteorological parameters, such as air temperature, cloud base height, relative humidity, wind speed, air pressure, azimuth angle, and zenith angle. Data analytic tools help to forecast output power of renewable systems.;In this thesis, data analytic tools are used to provide an increment in the share of renewable resources in production electricity. An artificial neural network (ANN) model was created to estimate hourly solar irradiation and wind speed. Dataset was recorded in Antalya in 2013 by the Turkish State and Meteorological Service. Furthermore, this study is purposed to improve the accuracy of energy forecasting. This improvement could be realized by finding the best structure of ANN for energy forecast, and by developing the performance of ANN using data analytics tools, such as Genetic Algorithm (GA) and Principal Component Analysis (PCA). Analysis of data can distinguish relevant information and extract useful knowledge from apparently unrelated data that is formed in a massive volume. This meaningful information can be used to forecast environmental behavior, which helps maximize the value of wind and solar output power estimations.;In addition, PCA is chosen for reducing the dimension of datasets to save time and memory cost with a better network performance. GA is chosen to improve network performance by finding and fixing the best weight for ANN. |
