Robots in substations: value only if they see what SCADA cannot

I'm a power systems engineer in Pittsburgh working on substation upgrades. We piloted a quadruped robot with thermal and acoustic sensors for yard inspections. It did fine finding hot spots and noisy bushings, but the real win was catching a slow SF6 loss on a 115 kV breaker that was not wired into SCADA. The early read let us plan a service window instead of a forced outage. Lesson one: make robots look at what SCADA cannot see.

Two surprises. First, connectivity. Yard WiFi was useless once you mix steel, fences, and RF noise. Private LTE and a simple waypoint plan beat any fancy autonomy. Second, workflow. Robot alerts landed outside our normal dispatch, so they were ignored. Once we mapped findings into the same work order system and standing alarms, people acted. Also, define hard no-go zones tied to grounds and holds so the robot never conflicts with switching.

For those using robots or heavier substation automation, where did it truly cut truck rolls or prevent an event? What integration step or hazard did you underestimate?