Trust, but Filter: Observability for a Battery Fleet When the Sensors Lie

Entrix trades grid-scale batteries across European power markets, and every decision depends on one number: the battery's state of charge (SoC). However, SoC telemetry data is noisy, lags, drifts, and occasionally spikes to values that are simply wrong, resulting in poor trades and lost revenue.

In this talk, Vasily Geyer, Principal Engineer, and Sebastian Kerz, Senior Data Scientist, will share how their teams at Entrix combined classic observability with data-driven prediction models to identify unreliable SoC values, track confidence in telemetry data, and make downstream decisions more robust.

Attendees will leave with a reusable pattern for monitoring systems where the data itself can’t be trusted, and with practical ideas for monitoring complex metrics pipelines and connecting observability to business outcomes.