ITEE Innovation Expo 2008 : Project Details

Investigation of Convergence for NEMMCO Monte Carlo Simulations

Successful long-term planning for system adequacy of the National Electricity Market (NEM) relies on the accurate estimation of key system indicators. These indicators include the expected energy not supplied to load points, also known as the Unserved Energy (USE). The complexity of the NEM, and the inherent uncertainty in factors such as generator availability, favours a stochastic Monte Carlo simulation method over an analytic evaluation in determining the expected value of these indicators.

A key challenge with Monte Carlo methods is the time required to complete each simulation of the NEM and the slow convergence to an expected value. To estimate USE for a specified year, the binomial sampling of each generator on the NEM and subsequent load flow calculation is required for each hour of that year. It follows that a better understanding of the factors affecting convergence could lead to further optimisation of the number of simulations required to obtain the desired confidence interval for the USE indicator.

While the sampling of generator availability suggested a binomial probability distribution function (pdf), the different generator sizes and forced outage rates (FORs), different number of generators for different hourly load levels, and the need for a reserve margin, distorted the sampling to a non-standard pdf. However, provided sample size is sufficiently large the Monte Carlo simulation results will approximate a normal distribution in accordance with the Central Limit Theorem, in which case, the sample mean and sample variance can be used to estimate convergence after each simulation.

This approach of analysing the stochastically sampled data can be applied more generally for a wide range of complex or inherently random processes.