A comparison of forecasts methods for the calculation of operating envelopes

Rising residential solar photovoltaics penetration in Australia challenges distribution companies to maintain voltage and thermal utilisation of low voltage networks within technical limits. Operating Envelopes (OEs)—time-varying limits at the customer connection point—are being implemented to mitigate these issues. However, the effectiveness of OEs depends on forecasts of power flow-related parameters, such as customer net demand and transformer voltages. This study compares the performance of four different approaches to calculate OEs—Asset Capacity, Asset Capacity & Critical Voltage, Asset Capacity & Delta Voltage, and Ideal—using different forecast methods: Persistence, Multiple Linear Regression, Random Forest, k-Nearest Neighbours, and an Ensemble. Results show that even with a perfect forecast, thermal and voltage issues were common in the Asset Capacity and Asset Capacity & Critical Voltage OEs, while the Asset Capacity & Delta Voltage and Ideal OEs avoided such problems, with only the Ideal OE fully utilising network assets. Moreover, forecast methods’ errors exacerbated thermal and voltage issues for most OEs. Nevertheless, the ensemble forecast method minimised them for the Ideal OE. However, for the three Asset Capacity OEs the ensemble forecast provided limited improvements over simpler forecasts like Persistence, making Persistence a more practical choice.