You are viewing content from a past/completed QCon

Presentation: Tesla Virtual Power Plant

Track: Architectures You've Always Wondered About

Location: Fleming, 3rd flr.

Duration: 4:10pm - 5:00pm

Day of week: Tuesday

Share this on:

This presentation is now available to view on InfoQ.com

Watch video with transcript

Abstract

A Virtual Power Plant (VPP) is a network of distributed energy-resources (often solar, wind, and batteries) that are aggregated to provide smarter and more flexible power generation, distribution, and availability. A VPP leverages assets for more than one purpose, and, in doing so, decentralizes generation, enables market participation for small generators, increases grid reliability, and smooths the intermittency of renewable generation while improving the economics of renewables.

Tesla's VPP consists of vertically integrated hardware and software, including both cloud and edge computing. The Tesla Energy infrastructure platform ingests and aggregates telemetry from tens of thousands of assets, including Powerwalls, Powerpacks, and Megapacks, with low latency, and robustly distributes control commands, handling measurement uncertainty and network intermittency. Tesla Autobidder leverages this infrastructure platform to optimize diverse asset fleets, decomposing the optimization algorithm across edge and cloud intelligence.

This talk will explore the evolution of Tesla's VPP architecture, including:

  • The use of distributed, actor-model programming for virtually representing physical assets and performing low-latency, hierarchical aggregations of telemetry.
  • Managing complex workflows for forecasting, optimization, bidding, and control that must be predictable, reliable, and resilient under market-timing, algorithmic, and physical constraints. These workflows must handle diverse inputs, including inputs from third parties, that are only eventually consistent.
  • Using Reactive Streams for reliably interfacing disparate systems, respecting resource constraints, and managing high-volume, near-real-time data streams.
  • Introducing functional programming techniques to make the software more composable, reliable, and testable.
  • Trade-offs between training and running global and local forecasting and optimization models.
  • The development of responsive user-interfaces for operations and energy-market participation.

Speaker: Colin Breck

Sr. Staff Software Engineer @Tesla

Colin Breck has experience developing software infrastructures for the near real-time monitoring and control of industrial applications. At Tesla, he works on distributed systems for the monitoring, aggregation, optimization, and control of distributed-energy assets, including solar generation, battery storage, and the Supercharging network. Previously, he worked on the PI System at OSIsoft, a time-series platform for industrial monitoring and automation. He is interested in the intersection of developing people, teams, and software systems. He writes a monthly essay at www.colinbreck.com.

Find Colin Breck at

Speaker: Percy Link

Staff Software Engineer @Tesla

Percy Link is a staff software engineer on the Energy Optimization team at Tesla, working on the Autobidder platform. Percy earned a Ph.D. in climate science and entered industry as a data scientist in smart grid applications. She has evolved into a software engineer, focusing especially on the real-time optimization and control of flexible energy resources.

Find Percy Link at

Last Year's Tracks