Machine Learning-Based Classification of Residential Solar PV Adoption Trends

The rapid growth of residential solar systems has impacted distribution grid operations, necessitating utilities to investigate the key factors driving solar adoption. This study has explored the key socio-economic and perception-based factors influencing residential solar adoption. Acquaintance with a solar PV owner, urban households, moderate income, home ownership, and perception of energy security encourage solar adoption. Perceptions like environmental benefits, energy savings, and confidence in solar installers distinguish potential adopters from non-adopters. The study further applies five classification models-Naive Bayes, Logistic Regression, Support Vector Machine (SVM), Random Forest (RF), and Multivariate Adaptive Regression Splines (MARS) to classify adopters and non-adopters and introduce adoption estimation rate as a performance metric. Naive Bayes and RF exhibit the best accuracy, around 80% accuracy in classification, while MARS and logistic regression consistently predict solar adoption rates close to the actual solar adoption rates.