AI-Enabled Power Management for Maximized Electricity Costs Savings

  • Conference Program
  • Hydrocarbon Processing
  • April 10, 2019
  • 10:15 am - 11:00 am

IBM and Plains Midstream Canada are developing real-time insights into power costs and consumption through the integration of the industrial IoT, machine learning and public domain data. The result is a multi-use power management solution that can:

Predict power prices. As a liberalized energy-only market, Alberta power prices suffer from both short- and mid-term volatility. In a single day, prices typically fluctuate between $25/MWh and $100/MWh. In response, IBM has developed a power price predictor that can optimize for pipeline scheduling and maintenance work orders. For each work order that is shifted to midday, when prices are high, companies can save up to 50% in electricity costs.

Give advance warning of price spikes. During power plant outages, electricity prices can spike by nearly 2,000% (to $1000/MWh), and last for hours. These spikes occur about 0.5% of the year, but may generate 7% of a plant’s power costs. IBM has developed an early warning system for short-term price spikes, enabling a plant operator to decide if equipment should be shut down based on the economics of the facility.

Support capital funding decisions. Machine learning also provides insight into power tradeoffs for capital project decision support, operations optimization of expenditures, and forecasting spend across facilities over the long term.