Long-term energy evaluation of PV systems that use micro-inverter configuration (microinverter PV systems) is currently unclear due to the lacking of sufficient longitudinal measurement data and appropriate analysis method. The poor knowledge about impact and aging of micro-inverter PV system affects the comprehension and accuracy of PV design and simulation tools. In this paper, we propose a machine learning approach based on the mixed-effect model to compare and evaluate the electrical energy yield of micro-inverter PV systems. The analyzed results using a 5-year period data of PV stations located at Concord, Massachusetts, USA showed that there is no significant difference in yearly electrical energy yield of micro-inverter PV systems under shading and non-shading condition. This finding has confirmed the advantage of micro-inverter PV system over the other PV types. In addition, our work is the first study that identified the average degradation rate of micro-inverter PV of 3% per year (95% confidence intervals: 2% − 4.3%). Our findings in this study have extended substantially the comprehensive understanding of micro-inverter PV system.INDEX TERMS Mixed-effect model, micro-inverter PV system, micro-inverter configuration, longitudinal measurement, fixed effects, random effects.