Master Degree Thesis Data Analysis and Modeling for Fault Detection in Solar Photovoltaic (PV) System- Mokpo National University Department of Electronics Engineering-Author Prasis - 광전지(PV) 시스템의 고장 탐지를 위한 테이터 분석 및 모델링South Korea

Prasis Poudel A thesis Submitted for Partial Fulfilment of the Requirements for the Degree of Master of Engineering Department of Electronics Engineering 
Graduate School Mokpo National University August, 2017
 광전지(PV) 시스템의 고장 탐지를 위한 데이터
분석 및 모델링

ABSTRACT
This thesis presents the solar (PV) Power output data analysis and modelling using least mean square (LMS), linear regression and multiple linear regression algorithms and comparison between them to find the best model for applying in the PV system Fault detection. This method has been developed and validated using climatic and electrical output obtained from a SANYO 200 Wp photovoltaic modules installed at the Hae-Nam, Korea. This modelling includes the correlation of solar PV power output and solar irradiation. In modelling algorithms, PV power is modelled adaptively as a function of solar irradiation and each model is compared in terms of estimated error performance based on statistical and graphical methods. From the result, it was found that the multiple linear regression modelling is the best for solar PV modelling with MSE 0.00818 with modelling error 1.58% which is less than that compared to the model using the least Mean square (LMS) having 1.97% and linear regression 5.98%. ii | P a g e After successfully modelled, the solar Photovoltaic (PV) power output as a function of solar irradiance, resulting best model is used for the development of practical fault detection. Our modelling results had fairly low complexity with high fault detection rates. The fault detection is based on the analysis of the power losses using the linear regression modelling. The model estimated by stepwise linearity of the PV power output as a function of irradiance. The results obtained from this modelling indicate that the under normal condition the solar radiation and PV power output have a very strong positive correlation and very useful for solar PV data prediction. In addition, the observations below the proposed linear functions are considered as the faulty PV data. From Overall results, we can conclude that this PV system data modelling and fault detection approach is reasonable for PV power estimation and faulty data analysis.
LINK
http://www.riss.kr/search/detail/DetailView.do?p_mat_type=be54d9b8bc7cdb09&control_no=6310cda13afffe46ffe0bdc3ef48d419