
Application of the mixed effects model for analysing photovoltaic datasets and interpreting into meaningful insights
Author(s) -
Le Ngoc Thien,
Asdornwised Widhyakorn,
Chaitusaney Surachai,
Benjapolakul Watit
Publication year - 2021
Publication title -
iet generation, transmission and distribution
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.92
H-Index - 110
eISSN - 1751-8695
pISSN - 1751-8687
DOI - 10.1049/gtd2.12160
Subject(s) - context (archaeology) , mixed model , variance (accounting) , computer science , statistics , photovoltaic system , raw data , time series , contrast (vision) , analysis of variance , series (stratigraphy) , regression analysis , linear regression , repeated measures design , linear model , econometrics , data mining , mathematics , artificial intelligence , engineering , geography , paleontology , accounting , archaeology , electrical engineering , business , biology
Energy power studies commonly apply the analysis of variance (ANOVA) method in data analysis context to find the significant difference among many compared groups or to evaluate the impact of an influential factor. However, most of the datasets in these studies are time series data or longitudinal data, which are collected from the same object over some periods. Therefore, violating the independent assumption of ANOVA is the error commonly made. This leads to the misinterpretation of the comparison tests. In this article, this problem is solved by providing the application of mixed effects model (MEM) as a viable alternative to the ANOVA. In detail, two models based on MEM are proposed to analyse the time series data of micro‐inverter PV stations located in Concord city, Massachusetts, USA. In the first scenario, the average models are implemented to compare the seasonal variation of monthly yield taking into account the effects of shading conditions and different orientations. In the second scenario, the linear regression model based on MEM is implemented to estimate and compare the decline rate in a 6‐year period of PV stations. The analysis results have shown that the models based on MEM performs better than ANOVA in interpreting raw time‐series dataset into meaningful insights.