PV Hosting Capacity Estimation in Low-Voltage Secondary Networks Using Statistical Properties of AMI Data

• Main Authors: Yusuf, Jubair, Azzolini, Joseph A., Reno, Matthew J.
• Format: Conference Proceeding
• Language: English
• Subjects: advanced metering infrastructure (AMI); Analytical models; Data models; data-driven analysis; hosting capacity; Low voltage; machine learning; Meters; Prediction algorithms; Reactive power; Sensitivity; smart meter data
• ISSN: 2643-8798
• DOI: 10.1109/ISGT-LA56058.2023.10328317

The widespread adoption of residential solar PV requires distribution system studies to ensure the addition of solar PV at a customer location does not violate the system constraints, which can be referred to as locational hosting capacity (HC). These model-based analyses are prone to error due to their dependencies on the accuracy of the system information. Model-free approaches to estimate the solar PV hosting capacity for a customer can be a good alternative to this approach as their accuracies do not depend on detailed system information. In this paper, an Adaptive Boosting (AdaBoost) algorithm is deployed to utilize the statistical properties (mean, minimum, maximum, and standard deviation) of the customer's historical data (real power, reactive power, voltage) as inputs to estimate the voltage-constrained PV HC for the customer. A baseline comparison approach is also built that utilizes just the maximum voltage of the customer to predict PV HC. The results show that the ensemble-based AdaBoost algorithm outperformed the proposed baseline approach. The developed methods are also compared and validated by existing state-of-the-art model-free PV HC estimation methods.