A comparison of inertia estimation in power networks hosting significant renewable energy

Abstract

In an ever-changing world of power generation, the increased use of renewable generation can pose significant problems for power quality, particularly by affecting a system's sensitivity to frequency deviations. Mechanical inertia is inherently available in a system utilising mechanical generation, this added benefit ensures frequency stability. However, in a system with increased renewable penetration, the amount of mechanical inertia might be insufficient, and knowing how much inertia is available for stability should be investigated. While various methods exist in the literature to estimate inertia, how accurate these methods are for estimating inertia in a system with significant renewable penetration was investigated. Evaluating available inertia enables power system operators to implement stability measures. These measures may include emulating inertia through static generation sources, utilising reserve capacity during contingencies, or deploying flywheels to provide additional mechanical inertia. In this study several inertia estimation methods from literature have been explored, including the swing equation with consecutive measurements, the swing equation applied to data windows, fifthorder polynomial fitting of frequency curves post-disturbance, electromechanical swing analysis using only active power measurements, and the R-V-RV method. To compare these methods, synthetic data were generated through computer simulations using three IEEE standard test systems: the 2-Zone, 9-Bus, and 14-Bus systems (converted to 50 Hz nominal frequency). Four disturbance events were instigated on each system: generation loss, a 3-phase fault, load decrease, and load increase. The generated data allowed for a comparative analysis of different inertia estimation methods. Findings from this work suggest that inertia estimation is highly sensitive to various variables, where improper data selection (hyperparameters) can lead to inaccuracies. Further findings suggest that no one method should be used for all cases, but rather, matching the system type and available event type data to its proposed method can deliver more accurate results. The estimated inertia constant at a given moment reflects the instability a system experienced at that moment. By plotting multiple estimation constants over a period throughout an event, a better picture is seen of the system's sensitivity to change in frequency during a sudden mismatch between generation and loading. From plotting the system stability of a real-life measured system over time, it became clear when exactly a sudden change in generation or loading was present and enabled estimation of the available inertia. System events were analysed using this approach, enabling an evaluation of system stability under different generation mixes, specifically comparing scenarios with high PV penetration to those where a single synchronous generator supplied the load. The results indicate that stability varies across different measurement locations and is largely influenced by the type of generation electrically closest to the measurement point. A relationship between higher PV penetration, relative to synchronous generation, and increased system instability was observed by plotting the estimated inertia over time under varying PV contribution levels. A lack of available inertia in a system to its required inertia may cause frequency runaways. Tracking inertia in a system is useful for grid operators to know when to deploy additional measures for improving the frequency response. This study found that estimating inertia can be done with accuracy, by plotting the stability index of a system over time.