System inertia and Rate of Change of Frequency (RoCoF) with increasing non-synchronous renewable energy penetration

Abstract

This paper explores the relationship between system inertia and Rate of Change of Frequency (RoCoF) in a changing world with increased penetration of non-synchronous renewable energy power generation (wind and solar PV). Research performed for the Irish Regulator (CER) showed that a RoCoF of 1 Hz/s measured over 500 ms can be tolerated by consumers and generators alike. Simplified dynamic studies determined that RoCoF of greater than 1 Hz/s can result in pole slip to selected generators. Measurement of the RoCoF also showed that 500 ms was the shortest period for the Irish system to avoid spurious trips due to inter-area oscillations. For South Africa, using conservative assumptions for RoCoF (1 Hz/s) and a slightly larger than expected credible multiple contingency, the minimum system inertia was determined. A production cost model was then used to solve the unit commitment and economic dispatch problem with hourly time resolution for 2030 and 2050 for a range of scenarios. Applying typical inertia constants for all generators, the system inertia for each hour was determined for each scenario. The minimum system inertia was then overlaid following which it was determined when there was insufficient system inertia and for how many hours. As expected, in relatively high non-synchronous generation scenarios, there was insufficient inertia by 2030 for a small number of hours of the year (˜5%) with worst-case inertia being ˜35% below minimum inertia while by 2050 there was insufficient system inertia for almost half of the year with worst-case inertia being 90% below minimum required inertia. Various technologies were presented to improve inertia (synchronous and synthetic) with the most expensive synchronous technology (rotating stabilisers) costed for the scenarios considered. In this regard, provision of additional system inertia was always <1% of total system costs. These static calculations were then supplemented by dynamic simulations in a System Frequency Model (SFM) of the South African network using the industry accepted DIgSILENT PowerFactory tool. Good alignment between static and dynamic calculations were found where the 1 Hz/s RoCoF was ensured when the required additional inertia was added as calculated. Future work needs to include an analysis of frequency stability under disturbances (frequency nadir, settling frequency, frequency restoration time) as well as frequency control under normal conditions and not just RoCoF to ensure these remain within acceptable limits. This will include aspects of additional reserve requirements (due to supply-side variability), the probability of larger contingencies on the system and impact on under frequency load shedding schemes.