.
Smart devices are essential to ensure the stability of the power grid and resilience to intermittent energy production. However, smart devices can also be the target of cyber adversaries that may exploit false data injection attacks (FDIAs) to induce unstable grid conditions. A practical consideration of FDIA mitigation approaches is addressed here: given a finite available budget, for which smart device should cyber-threat mitigation be deployed first? In this work, this question is answered by identifying the so-called most-sensitive devices, i.e., the devices that, if compromised, can let an adversary induce the most serious grid instabilities. The method proposed utilizes an adversarial reinforcement learning (RL) framework to identify the k-most-sensitive smart devices (here, smart inverters). The adversarial agent can tamper with the compromised inverters’ active and reactive operating power setup points, with the goal of maximizing voltage deviations. Numerical results show that the proposed RL method finds the optimal attack scenarios for 1-point failure and the near-optimal solution for the 2-point case. Additionally, the proposed RL method achieves an 8.8× speed-up ratio in running time compared to the brute force method for the 2-point case.
Abstract A large comparison exercise has been performed featuring aerodynamic and aero-elastic simulation cases on the IEA 15MW reference wind turbine in various conditions, containing results of 30 codes ranging from BEM to CFD. More than 10 different variable types ranging from lifting line variables to pressures, loads and velocities have been compared for the different conditions, resulting in many comparison plots. The result is a unique insight in the current status and accuracy of rotor aerodynamic modeling. Although there are no measurements on this turbine, mutual comparison of model results provided useful insights into the performance of rotor aerodynamic models. Preparatory simulations on the 15MW RWT at constant uniform conditions generally showed reasonable agreement in the aerodynamic response between engineering and higher-fidelity models, provided the turbine was considered rigid. However, including flexibility effects led to more discrepancies, largely due to differences in blade torsion, which in turn impacts the aerodynamics. Even at very moderate wind speeds the blade tip torsion angle could be in the order of 2 degrees where large differences were found between the partners results. Following the preparatory cases, simulations under turbulent conditions were performed. Several turbulent boxes were generated using high-fidelity CFD models and the results were compared mutually. Some differences appeared in the turbulent boxes, which could be expected from convection differences. At first sight the differences seemed small. However, when the boxes were fed into an aero-elastic code, the differences became significant enough to affect load response. When supplying the sampled wind speeds from the turbulent box as input to engineering-fidelity models, it was interesting to find that these models showed a much higher standard deviation in loads compared to higher-fidelity models. This confirms the finding from previous numerical studies that engineering models tend to overpredict fatigue loads, also for a large sized rotor.
This paper presents ongoing work to enhance primary frequency response on the Puerto Rico grid with a practical methodology and software toolkit to monitor generator frequency response performance and validate governor models using SCADA measurements. The automated pipeline includes event detection; data extraction and conditioning; frequency-response analysis with computation of the frequency response measure (FRM) to classify supportive versus adverse behavior; rapid droop estimation via a simplified characteristic; and in-depth, disturbance-based model validation using a PSS®E play-in. Benchmarking indicates that a simplified droop method agrees sufficiently with full-model simulations to enable rapid screening prior to detailed validation. Application across multiple units and events identifies disabled or mistuned governors, deadband effects, and needs for parameter calibration and model updates. Results show improvements in fleet frequency response relative to prior years, despite persistent variability in unit response and gaps in meeting performance requirements for some generators. The approach complements required staged testing and supports continuous fleet-performance monitoring.
