Introduction

Household electricity consumption is a significant part of total energy consumption. Trends in the spread of remote work and distance learning only strengthen this contribution. Modern smart grid technologies allow for detailed analysis of the consumption patterns of each individual household. However, the task of analyzing the time series in aggregate, comparing, and classifying households is not easy, since each such series is unique. Modelling the stability of consumption using time series of readings is the main subject of this presentation. We present our method for monitoring the stability of residential electricity consumption.

Methods
As a measure of stability, we use the auto-similarity coefficient defined as the geometric mean of pairwise correlations between fragments (windows) of the corresponding time series. The method was introduced in our previous work. Here, we test the applicability of this approach to a real-world data set.

Results
This study found that one week is an appropriate window size for studying the stability of consumption. And also the capabilities of the method are demonstrated for real data of selected Swedish households. The method also reveals seasonal differences; for example, with a high stability of the pattern in the winter months, the same household has low stability in the summer vacation period. Cases with both a high degree of stability and low stability indicators are considered.

Conclusion

The proposed method can be applied to the analysis of the stability of electricity consumption and thus enriches the arsenal of mathematical modeling methods.