On Possibilities of Using Smart Meters for Compulsory Load Shedding Supported by Load Forecasting

Doctoral thesis, 2018

The smart meter rollout is progressing in several parts of the world with early adoption in some parts, e.g., Europe. Remote ON/OFF control switch of the smart meter allows distribution system operators to switch the smart meter of any customer remotely. This thesis investigated the possibility of using ON/OFF control switch of the smart meters for compulsory load shedding supported by load forecasting.

Acute situations, e.g., critical power shortage could require today compulsory load shedding as a last resort if the power reserve becomes insufficient. The compulsory load shedding is typically done from medium voltage substation level, and in that case, all customers including emergency service providers located under the affected substation would lose power. By using the remote ON/OFF control switch, it is possible to exclude the vulnerable groups of customers such as elderly and also socially critical customers such as clinics, pharmacies, and fire stations.

Three field tests have been performed on small-scale load shedding using the smart meters. The results have shown that the smart meters’ switching has no or negligible impact on the power quality at the low voltage level of the grid. Moreover, existing challenges in the present smart metering system, e.g., the reliability of confirmation report on smart meters’ switch status, are identified. Thus, demands that need to be put on the future smart metering system are identified.

This thesis developed large-scale smart meters’ switching model based on the field tests’ results. Moreover, load forecasting models are developed using Artificial Neural Network method to forecast load at the individual customer level and also at low aggregation levels, e.g., low voltage substation level.

The results from aggregated load forecasting models, e.g., at an hour with high load condition, have shown that the total load in, e.g., a 10kV residential grid can be estimated with an error of around ± 3% by using up to previous day’s hourly smart meter data as input predictor. Moreover, the results from a simulated example of selecting low voltage areas for load shedding have shown that, compared to load estimation from average load values, aggregated load forecasting models could help to save around 25% of number of customers from unnecessary load shedding.

The simulation results on voltage calculation at each of the low voltage substation during and after the load shedding show that the voltage calculation using individual customers’ forecasted load values gives a negligible error (around ± 0.001 per unit), compared to voltage calculation using actual load values.

The smart meters can be used for the compulsory load shedding for excluding prioritized customers. However, assurance of accurate status update report of smart meters’ switch and consideration of delays in smart meters’ disconnection are required to perform compulsory load shedding within the allowable time. Moreover, the assurance of either more extended battery backup time for smart meters and meter data collection units, or quick communication network buildup capability, are recommended to enable rapid reconnection of smart meters if power system fails after compulsory load shedding.