Members of the Laboratory for Open Systems and Networks (E5) at the Jožef Stefan Institute published the paper "Performance of missing data imputation methods for electricity consumption time series: a comparative study" by Ramanpreet Kaur and Dušan Gabrijelčič in Electric Power Systems Research (2027).
Electricity consumption data collected from advanced metering infrastructure (AMI) often contain missing values due to measurement errors, communication failures, or device outages. In the paper, the authors conducted a comparative analysis of eleven widely used imputation methods – including statistical, time-series-based, and machine learning approaches – using high-resolution electricity consumption data from Slovenian households. The methods were evaluated using normalized root mean square error (NRMSE) as the key performance metric, and statistical significance testing was used to determine whether performance differences between leading methods are consistent across households and experimental settings. The results show that imputation performance is strongly influenced by both the missingness rate and gap size: interpolation performs well for very short gaps, while methods capturing historical and contextual consumption patterns perform better for medium and long gaps. Among the eleven evaluated methods, XGBoost achieved the best overall performance with the lowest average NRMSE, confirmed by pairwise Wilcoxon signed-rank tests. The consumer-level analysis also shows that the most accurate method on average is not always the most consistent across households.
https://doi.org/10.1016/j.epsr.2026.113963