Performance analysis and forecast of power efficiency of grid-connected solar power plant

In this thesis, parametric and nonparametric regression models have been developed for real-time monitoring and estimation of the power generation of Şenyurt (Erzurum/Turkey) Solar Power Plant (SPP) with direct grid connect and three-phase central inverter topology. In the developed regression based power prediction models, the data of the weather station sensors of Şenyurt SPP were used. The weather station within Şenyurt SPP has a solar radiation sensor, Photovoltaic (PV) cell temperature sensor, ambient temperature sensor and wind speed sensor. In the power generation forecast model of GES, active power generation values were used together with all weather station sensor data for 2020. Multiple Linear Regression (MLR) analysis, which is one of the parametric statistical techniques, was firstly used in the power generation estimation model. As in other parametric analyses, basic assumptions such as sufficient sample size, normal distribution of data, multiple linear correlation, singularity, and extreme value were examined in detail in MLR. K-Nearest Neighborhood Regression (K-NNR) and Bagging Regression (BR) models, which are non-parametric statistical techniques, were used secondly in the power generation forecasting model. The performances of the developed regression models were discussed in detail in the thesis. As a result, thanks to the estimation of the power produced in the SPP, the operators will be able to monitor whether they are producing an efficient and sustainable energy.