Large steel manufacturing plant

Energy optimisation at a large steel manufacturing plant

Design and implementation of an energy balance optimization software.

Steel production processes consume  vast amounts of energy, mainly as coal, oil-based fuels, gases, electricity and steam. They also count as by-products some considerably rich materials, such as gases from the coke plants or, to a lesser extent, from blast furnaces. Many steel plants have installed on-site power plants re-using those by-products and other fuels. However, as major actors on the grid, they also have a commitment with respect to the Transmission System Operator (TSO), to whom they need to supply a quarter-hourly balance planning. Moreover, with the growth of the demand, the carbon cost and the variable proportion of renewables in the energy mix, electricity pricing has dramatically changed; not only increasing, but also becoming more volatile.


With an existing – but old – power plant on site, consisting mainly of steam generators feeding steam turbines, the goal of the client was two-fold:

Provide controllers with real-time guidance on the use of the power plant components, so that the quarter-hourly energy commitments could be respected and costly imbalance penalties avoided. Optimize the weekly commitments to take best advantage of daily electricity price peaks and lows in a realistic machine utilization plan.


A necessary step to attaining the client’s target was to properly model each of the plant components, to account for their consumptions, start up / shutdown patterns, functional limits, yields … Based on the client’s expertise and thorough analysis of historical data, we then defined a flexible model ready for optimization.

The first goal was then implemented as a real-time module based on this model. The model is fed with functional data as well as real-time data fetched from sensors spread over the steel and power plants.  It is scheduled to run every minute to provide guidance for the remaining minutes of the current commitment period of a quarter hour.

The second goal also made use of the optimization model, but at a weekly level and with a precision of 15-minute periods. Equipment utilization takes into account their sometimes long startup/shutdown periods to best leverage the different sources of energy. For both modules, we worked in close collaboration with the client  to interface with existing databases and, more importantly, to provide relevant information to the users. This was especially critical for the real-time module, as output had to be used within the next few minutes!

The third goal was to assess the extent to which the identified flexibility can be leveraged on the reserve markets, and what the optimum strategy is towards bidding for primary, secondary and tertiary reserves.

Experience highlighted the importance of pre-treating the real-time data. Indeed, due to the distance, bias, age, process perturbation, synchronization errors etc, sensors often provide data that is not entirely accurate, or that would even violate the fundamental laws of physics if used directly!  We defined a pre-treatment module that corrects real-time data to respect the physics of the system and account for reliability divergence in each sensor.


Reduced energy bill

By taking the price of energy into account within the production process, the client could adapt energy-intensive industrial operations to minimize costs, while meeting their production targets.

Minimized imbalance costs

The real-time module enabled the client to instantly identify useful corrective actions in order to meet their quarter-hourly balance, or take advantage of opportunities on the balancing market.

What do we offer?

Our services