
Machine Learning Application Cuts 25% of Electricity Costs at Metallurgical Complex
THE PEAK LOAD PROBLEM AND ENERGY PENALTIES The client is a metallurgical corporation owning a large-scale steel rolling mill complex. A characteristic of the metallurgical industry is its vital dependence on massive electric arc furnace systems.
According to the power purchase agreement with the national electricity authority, the plant is subject to a multi-hourly pricing mechanism (Normal Hours, Off-Peak Hours, and Peak Hours). Furthermore, if electricity consumption in any 15-minute cycle exceeds the registered capacity limit (Peak Demand), the business will incur extremely heavy progressive penalties. Because the operation of the steel smelting furnaces relies entirely on the foremen's intuitive experience, the plant constantly switches on all heavy machinery simultaneously during peak hours, resulting in monthly electricity bills soaring to millions of dollars, severely eroding profit margins.
ESTABLISHING A FORECASTING SYSTEM USING MACHINE LEARNING ALGORITHMS The energy solutions team has comprehensively digitized the internal power grid and implemented a Smart Energy Management model based on Machine Learning algorithms:
- IoT sensor grid: Installing hundreds of smart electricity meters at each cable branch, each transformer, and each steel furnace. Current, voltage, and power consumption data are transmitted to the Edge Server system every second.
- Training a Machine Learning (ML) Predictive Model: The AI system is provided with a massive amount of historical data, including: production schedules, specifications of each steel batch, ambient temperature (outdoor weather affects heat dissipation performance), and electricity price cycles. The Machine Learning model uses regression algorithms to identify consumption patterns.
- Automatic Load Shifting System: Instead of simply issuing a deadlock report (after an event has occurred), AI is capable of extremely accurately predicting the total power consumption of the entire plant for the next 30 minutes. If the AI detects that the load curve is about to reach its peak, the system will automatically send an alert to the central control room. More sophisticatedly, the AI will directly intervene and temporarily reduce (or interrupt) the load on non-essential systems (such as auxiliary ventilation systems, idle cooling water pumping stations) for 10 minutes, "holding" the total power below the penalty level without disrupting the main steelmaking line.
COMPLETE OPTIMIZATION OF OPERATING COST STRUCTURE: After only 3 months of "learning" and process optimization, the AI system has delivered groundbreaking financial results:
- Eliminate 100% of the risk of electricity penalties: The plant no longer violates any peak power limits thanks to the predictive load management and cutoff capabilities of the forecasting algorithm.
- Reduce total energy costs by 25%: The AI system automatically recommends scheduling the most energy-intensive heavy steel batches for off-peak hours (after 10 PM), maximizing the advantage of lower electricity prices.
Increased equipment lifespan: Thanks to the stable current supply, avoiding sudden voltage drops due to overload operation, the maintenance costs of the factory's transformer system are significantly reduced.
Need a reliable technology partner for your business? See DUDI's dedicated tech team & white label.