2026 Event Schedule

Time-series data based (hourly or 8760 data) forecasting and power flow analysis have been an aspiration of electric utility companies for decades.  Many companies are stuck with coarse grain forecast and peak value (once a year) based power flow analysis due to outdated legacy system design, despite the tremendous advance in technology over the past decade.  Tasks that were once unthinkable, such as processing trillions of time-series data in hours, are in the realm of possibility now with recent advances in technology such as cloud-native design and open-source innovations.

Southern California Edison (SCE) embarked on a journey to turn system-wide time-series data (8760) based transformer structure level forecast and power flow analysis into reality, replacing outdated peak-based tools with a high performance, flexible, scalable, cloud-native, and open-source platform to modernize electric system planning. By utilizing massively parallel processing available on Snowflake for maximum performance, SCE achieved unprecedented results—processing 4.3 trillion load and DER records in 21 hours. 

SCE’s solution is a first-of-its-kind in the U.S. electric utility industry, delivering capabilities that no off-the-shelf platform can match. Unlike traditional tools that rely on peak-based analysis, it enables full time-series analysis down to the transformer structure level (meter level is also possible), meeting regulatory requirements that other systems cannot support. It is designed to handle multi-scenario forecasting, automated simulations at scale, and real-time investment creation. Its physics-based digital twin modeling, cloud-native architecture, and open-source design make it future-proof, eliminating hardware constraints while vastly improving grid planning efficiency with near-linear scalability.

The business aspiration and architecture design vision to turn it into reality, completely transformed the Engineering and Planning process at SCE. For example:

• It transformed engineering and planning by replacing a labor-intensive, error-prone, peak-value approach with an automated, fast, accurate, time-series-based methodology.
• It revolutionized architecture from a closed, desktop-based system to a cloud-native, open-source, data-driven design, enabling high performance, scalability, interoperability, extensibility, availability, and resilience.
• It strengthened data quality by centralizing model management and streamlining oversight on corrections crowdsourced across all units.
• It introduced a foundation for AI/ML-driven, physics-informed system planning, enabling automated bridging solutions and conversational-style violation resolution for the next generation of power system engineers.

The outcome is a significant advancement.  System planning effort has been reduced from months to hours, at a much lower cost, and with fewer resources

• Forecast generation for multiple scenarios, each outputting 10 years of hourly time-series values:
  • Processed 4.3 trillion records in 21 hours—30x faster than legacy system.
  • 95% reduction in data variance.
  • 90% reduction in operational costs.
• Territory-wide power flow analysis for 1 year of hourly timeseries (8760) data:
  • Processed 36 billion records in 1.5 hours.
• Territory-wide power flow analysis for 10 years of hourly time-series (8760) data:
  • Processed 360 billion records in 5 hours.
  • Legacy solution was simply not capable to process such workload

Session Sponsored by Itron