An approach for failure detection and diagnosis in photovoltaic power systems based on Weibull distribution function

Abstract

This paper proposes a method to predict photovoltaic (PV) 
energy production based on the hypothesis that during the 
hours of activity of the plant, its production behaves as a 
stochastic process with a Weibull distribution function of two 
parameters which take into account the characteristics of the 
photovoltaic plant and the weather parameters measured in 
real time by a sensor network. The versatility and accuracy of 
the model has been validated through the deployment of a 
real-time SCADA system. The system developed includes power 
generation prediction and real-time comparison between the 
power expected and the power measured, for failure detection. 
With the implementation of the proposed system was achieved 
a production improvement of more than 7% with respect to the 
periods with failures.