Simulation testing and improvements to network load modelling for a distribution network planning application

The work presented in this paper is part of a larger initiative to develop a Geo-based Load Forecasting (GLF) methodology for distribution master planning in South Africa. Part of the planning process, is forecasting the peak load for a geographical area with a given load class composition. This forecast is used to plan the capacity and timeline of infrastructure to service that area. In this paper, an overview of the GLF peak load estimation technique is provided as background. A Monte Carlo simulation based approach to evaluate the accuracy of this method was developed and is presented. Using the simulation, an evaluation of the current method and load library was completed and the results show that the current method significantly under-estimates the peak load. This can be attributed to changes in the load profile shapes of customers since 2007 when the current load profile library was derived, as well as simplifications in the derivation of the customer load behaviour. Specifically, the differences between the load behaviours of customers within a specific load class is ignored. An enhancement to the geo-based method is presented and compares well with simulated results across different customer classes and load aggregation periods. This improvement reflects the customer load behaviour more accurately and can readily be incorporated as part of the distribution master planning process.