Accurate labeling of phase connectivity in electrical distribution systems is important for maintenance and operations but is often erroneous or missing. In my project, we presented a process to identify which smart meters must be in the same phase using a statistical clustering method on voltage time series data. The data set was the hourly voltage of ~2000 smart meters across California over a 3-month period. To improve accuracy, we compress the data and reduce the size by using Fourier series – ultimately, we used an approximation that was 2% in size but still retained 80% of the original features.
