Impact of Partial Unplanned Outage Modeling Assumptions on Long-Term Capacity Planning Validation

Short-term power system models can be used to validate long-term expansion plans or to investigate the adequacy of possible future electrical power systems. When building these models, decisions need to be made about the level of detail that should be included. These decisions can be influenced by run time constraints as well as the availability of information. This paper investigates the impact that including partial outages has on the results of short-term validation models. The ratio between partial and full unplanned outages of generators is unique to the specific system being modelled. Projections on these ratios are not always available and the impact of simplifying the representation of outages by, for example, omitting partial outages is not documented in the literature. The 2019 IRP projection of the South African power system in 2030 is used as a case study to investigate the impact that modelling partial outages has on system reliability metrics and the capacity factors for the various technologies in the system. About 60% of all electricity generated in this weakly interconnected system is produced from coal fired power stations that experience partial outages, rendering this a useful case study. The system is modelled with varying ratios of full and partial unplanned outages. It is found that as the ratio of partial outages increases, the amount of electricity produced by coal fired power plants rather than other sources slightly increases, storage utilization significantly increases, and electricity production from gas and diesel plant significantly decreases. When going from a model with a 50-50 breakdown between full and partial unplanned outages to one with only full unplanned outages, storage utilization decreases by 40% and gas and diesel utilization increases by up to 44%. The amount of unserved energy and the loss of load probability decreases as the ratio of partial to full outages increases. Given the potential impacts of incorrect partial outage assumptions, the paper concludes that it is advisable to test the sensitivity of modelled results to the inclusion of partial outages in cases where information on their prevalence is not available. The impact of varying the relative size and duration of partial outages is found to have a much less significant impact on model outcomes. These aspects thus lend themselves to simplification. The findings presented in this paper are relevant to those engaging in long-term capacity planning and research informing such planning, especially for weakly interconnected power systems that are currently heavily dependent on generating technologies that are prone to partial outages.