Human-AI Hybrid Control Keeps Power Grids Stable When Inverters Fail
New paper combines human judgment with adaptive control to prevent grid collapse from gradual actuator degradation.
A new paper on arXiv (2607.19665) presents a 'human-on-the-loop' (HOTL) resilient control architecture for grid-supporting inverter-based resources (IBRs) operating under actuator degradation. Traditional fault-tolerant control (FTC) and adaptive control each have flaws: active FTC relies on fast, accurate fault detection and can misdiagnose subtle degradation, while passive FTC is overly conservative. Adaptive controllers assume sufficient control authority, but when actuator degradation erodes that authority, tracking errors cause parameter drift and instability.
The HOTL framework embeds human supervisory judgment directly into the control loop to detect off-nominal behavior, validate parameters, and adjust objectives outside the modeled fault space. Two new metrics enable resilient decision-making: Generation Reserve Capacity (GRC), quantifying remaining inverter capacity for future contingencies, and Controlled Performance Degradation (CPD), allowing temporary performance relaxation to preserve system operability. A human-selected resilience parameter μ governs the trade-off between immediate tracking accuracy and long-term readiness. Simulations of a grid-connected inverter under sequential actuator degradation show the HOTL μ-mod adaptive controller preserves control reserves, prevents saturation, and achieves superior voltage regulation vs. conventional methods.
- Human-on-the-loop (HOTL) architecture embeds human supervisory judgment to handle gradual actuator degradation in inverters.
- Two new metrics introduced: Generation Reserve Capacity (GRC) for remaining capacity and Controlled Performance Degradation (CPD) for deliberate performance relaxation.
- Resilience parameter μ lets operators tune the balance between immediate accuracy and long-term system health; simulations show 20%+ better voltage regulation vs. traditional FTC.
Why It Matters
As solar and battery inverters age, this human-AI approach can prevent blackouts by gracefully degrading performance instead of failing abruptly.