Contact form

Contact Form (Conditions: Gas, electricity, phone)

Contactdata/ General interest (Contact-Flap)
What is your interest?
Product-Section (Contact-Flap)
Gas supply informations (Contact Flap Options)
Your annual consumption
Your sites
Electrcity supply informations (Contact Flap Options)
Your annual electricity consumption
Your sites
Submit (Contact-Flap)

Reducing the costs of balance energy – An example from the steel industry

Balance energy is a key cost factor on steelworks’ electricity bills. Based on artificial intelligence, we have developed a digital solution that predicts short-term energy consumption more precisely and reduces the amount of balance energy for steelworks by up to 30%.

High electricity consumption at steelworks

Energy-intensive operations like steelworks have to bear high electricity costs. In the case of Swiss Steel AG in Emmenbrücke, electricity costs make up as much as 74 per cent of the total energy costs. In addition, the uncertain and fluctuating price trends pertaining to the costs of balance energy represent a financial risk for steelworks.

For this reason, we and our partner Swiss Steel have developed a solution on the basis of artificial intelligence and real-time IoT incorporation, which predicts energy consumption more precisely and reduces the amount of balance energy significantly.

Good to know

The cost of balance energy is calculated on the basis of the difference between the scheduled (i.e. predicted) and actual electricity consumption within a 15-minute interval. For steelworks with an electric arc furnace, precise prediction of energy consumption is particularly difficult, as the exact furnace activation time cannot be scheduled precisely. For this reason, an average consumption profile is generally determined on the preceding day (day-ahead nomination). The greater the deviation from this nomination, the greater the additional costs charged by the energy supplier.

Prior announcement of balance energy at the right time

Steelworks normally announce their consumption profile to the energy supplier one day in advance (day-ahead nomination). At that time, the actual consumption can only be estimated very roughly. Short-term prediction with greater precision would be better. This is why we have developed an algorithm to predict energy consumption on the basis of real-time data for Swiss Steel at the Emmenbrücke steelworks. The prediction announcement’s lead time is now no longer one day, but just a few minutes before the time of acquisition.

Our quarter-hourly predictions correct the original electricity schedule and are forwarded to the energy supplier, who can thus trade the difference in volume on the short-term (intraday) energy market or compensate for it with flexible power stations.

Specifically, our solution includes the following elements:

Real-time data processing
Capture of all relevant data in real time for calculation of the consumption prediction and consolidation on the Alpiq Energy AI platform.

Prediction algorithm
A neural network that was trained using historical consumption data and production schedules, and can comprehend how they interrelate, first analyses the continually captured real-time data, then processes it together with the current production schedule.

Automated intraday renomination
The prediction algorithm is executed every 15 minutes and the prediction serves as a corrected schedule for the originally nominated consumption.

The result: 17% less balance energy costs

The continually updated predictions enable Swiss Steel to convey short-term deviations in energy requirements to the electricity supplier and to nominate them on the intraday market. Consideration of additional process data, scheduling parameters and nomination behaviour has enabled us to gradually reduce the costs of balance energy by 17% since the project began in July 2018. With constant optimisation of these parameters, we expect further savings and a reduction in balance energy of up to 30%.

The benefits at a glance

  • Steelworks reduce acquisition of balance energy and the associated costs.

  • Steelworks decrease the risk of unforeseeable price peaks.

  • Steelworks gain more insight into load behaviour and its influencing factors.

  • Steelworks with full supply obtain a better negotiating position when purchasing energy (reduction of risk premium for bad predictions).

  • The modular structure of the Alpiq Energy AI platform enables other measures, e.g. incorporation of an industrial load management system for additional reduction of load peaks and grid acquisition costs.

  • The algorithm is easy to integrate into existing energy management systems.

  • Steelworks can manually adapt the automatic prediction, if necessary for operational reasons.

Further references