Title:	Worldwide Moisture Ingress Evaluation in Photovoltaic Modules Using Finite Element Simulations and Machine Learning
Author(s):	S. Mitterhofer, J. Ascencio-Vasquez, X. Gu
Keywords:	Moisture Ingress, Finite Element Method, Machine Learning
Topic:	Photovoltaic Modules and BoS Components
Subtopic:	Materials for PV Modules, Durability, Reliability and Accelerated Testing Methods
Event:	8th World Conference on Photovoltaic Energy Conversion
Session:	3EO.1.3
Pages:	609 - 613
ISBN:	3-936338-86-8
Paper DOI:	10.4229/WCPEC-82022-3EO.1.3
Price:	0,00 EUR
Document(s):	paper

Abstract/Summary:

---

Ambient water can be absorbed by photovoltaic (PV) modules, affecting their degradation. This paper presents a two-step approach to evaluate the moisture profile inside fielded PV modules around the world. The first step consists of moisture ingress simulations at more than 17000 global locations using a cluster, running up to 160 simulations in parallel. Results of the simulations show a low R2-value of 0.65 when correlating the ambient relative humidity and the saturation moisture concentration in front of the cell, emphasizing the need for such simulations to provide more accurate inputs for degradation modelling. In the second step, we train a machine learning algorithm to obtain a model finding a better correlation between the important climatic parameters and the simulation results. The ML algorithm can accurately find such a correlation, shown by the R2-value of 0.98. We use the algorithm to further predict water ingress and interpolate the results to other global locations in a higher spatial resolution of the climatic input parameters.