Case Study: 4.1% Energy Reduction at an EU Electric Steel Plant

In early 2024, a major European EAF steelmaker partnered with PRAX to optimize energy consumption across their melt shop. The plant produces over 1.5 million tonnes annually and operates three electric arc furnaces around the clock.

The goal was straightforward: reduce electrical energy per tonne of liquid steel without changing product mix, raw material suppliers, or production throughput.

Baseline and Approach

We began with a 90-day observational period, instrumenting the existing data infrastructure — Level 2 automation, power metering, weigh-bridge records, and spectrometer logs — to build a unified heat-by-heat dataset.

The baseline energy consumption averaged 412 kWh/t liquid steel. Variation was high: the interquartile range spanned 38 kWh/t, driven by differences in scrap mix, tap-to-tap time, and operator practice across three shifts.

Our approach focused on three levers:

1. Charge optimization — adjusting bucket composition to reduce average melting difficulty while maintaining chemistry targets.
2. Power profile tuning — recommending transformer tap and electrode regulation setpoints based on predicted scrap behavior.
3. Oxygen/carbon injection timing — coordinating chemical energy input with the arc phase to reduce total electrical demand.

Results

Over a six-month deployment period, average energy consumption fell to 395 kWh/t — a 4.1% reduction. Critically, the improvement was consistent across all three furnaces and all shift patterns, suggesting the gains came from systematic recipe and setpoint changes rather than individual operator behavior.

The interquartile range also narrowed to 26 kWh/t, indicating more consistent operations overall.

Economic Impact

At the plant's electricity cost of approximately EUR 85/MWh, the energy savings translate to roughly EUR 1.4/t liquid steel. For a plant producing 1.5 Mt/year, this represents annual savings of approximately EUR 2.1 million — well above the cost of the PRAX deployment.

Electrode consumption also decreased by 3.8%, likely a secondary effect of shorter power-on times and more stable arc conditions.

Lessons Learned

The largest single contributor to energy savings was charge optimization — not because the models were sophisticated, but because the existing practice relied on static recipes that hadn't been recalibrated in over two years. The data infrastructure PRAX installed made recipe drift visible for the first time.