A new data-driven method based on Niching Genetic Algorithms for phase and substation identification

Phase topology identification in low-voltage distribution networks: A Bayesian approach

“…This major drawback makes these approaches unideal. Nevertheless, as shown in [19,26,27] and validated by this study's results, the correlation between variations in consumer consumption and supply in phase feeders can solve the drawback of the energy-balance approach. Thus, the proposed method does not use the energy conservation law (Eq.…”

“…As it was said, after the first iteration, a change in the process is included to increase the convergence. As proven in [19,20,26,27], the correlation of customer variations with its phase increases if other customers' consumption is subtracted from their phases. Therefore, removing some of the customers' consumption in the phase with the highest probability could improve convergence if the hypothesis with the highest Fig.…”

“…These changes result in a significant improvement in accuracy with fewer samples and a lower percentage of missing measurements than the initial algorithm. The same authors also published a novel phase identification method based on the same ideas, but using genetic algorithms [27]. This method obtained better results than the previous one, but only when the percentage of measured energy is near 100 %.…”

“…This is similar to the high-frequency part of the load profile signal in the frequency domain. With these variation signals, and considering a customer n connected to a phase p, other related studies [19,20,26,27] have shown that there is a strong correlation between the customer variation profile (VC n ) and the aggregated variation profile at the feeder of the phase in which it is connected (VF p ). A simple example to understand the rationale behind this idea is the next: if a customer has a significant change in their consumption pattern (either reducing or increasing its consumption or generation), that change will be reflected in some way in its phase feeder.…”

“…In this study, the new evidence is the energy consumption profile of a customer, specifically its consumption variations, as shown in the previous section Xu et al [19] proved that the customer's consumption variations show a high correlation factor with its phase feeder consumption variations using the Pearson Correlation Coefficient (PCC). As mentioned in the state-of-the-art revision, this idea of using the PCC was also used by [20,26,27] in their phase identification methods, which showed good results. Thus, a possible way to establish the likelihood of the new data for each hypothesis (three in our case) is to perform the PCC between the variations of a customer and the feeder associated to that hypothesis.…”