Windscape.ai increases production and decreases maintenance with alerts of gusts and wind changes using AI data correlation on data from a scattered array of ultra-low-cost,

ground-mounted sensors.

 

 

Our experienced team is completing a field pilot to prove generation of actionable wind event alerts from our sensor data.

The Industry Context

  • 65,000 wind turbines installed in the US will generate 400 GWh in 2021

 

  • 400,000 wind turbines installed worldwide

 

  • $100B annual US and $600B world annual wind industry with 20% annual growth

 

  • Accelerating national attention to green energy.

 

  • Meanwhile 2018 US average wind PPAs were below 2¢/kWh - squeezing margins

 

  • Small production increases have huge impacts on margins

 

  • Turbine size and cost keeps growing - going to 6 MW onshore

 

  • Hundreds of thousands of dollars to replace a gearbox

Technology Innovation 

The vast majority of wind turbines today adjust after winds arrive causing lost energy production and exposure to damaging gusts. Available methods for providing wind event alerts are very expensive, and not widely deployed. 

 

Windscape.ai will be a 10x breakthrough in cost and effectiveness in generating 10-60 second alerts of wind changes and gusts. This will allow operators to increase wind farm energy production and turbine life by optimizing pitch and yaw settings for energy capture and protecting gearboxes and blades from damaging gusts. 

 

Windscape.ai uses a low-cost wifi mesh network of sensors across and surrounding the wind farm that feed into AI data correlation. This approach is 10x cheaper than any alternative. It is patented and made possible now through advances in edge AI and sensor-on-a-chip tech. 

Key Points

  • $100M addressable US market; $600M worldwide

 

  • Enabling wind turbine pitch and yaw ahead of wind events with 10 to 45 seconds lead time alerts

 

  • AI pattern recognition of telltale atmospheric signals

 

  • Ultra-low cost, robust hardware

 

  • Bottom-line benefits:

           Increased annual energy

           Lower O&M 

 

  • No turbine contact (no climbing, mounting,  or interfering with turbines or operations)
     
  • Enabled by advancements in edge AI and micro sensors 
     
  • Large or small wind farms

 

  • Self adapting to any land-base wind farm
     
  • IP patented and pat. pend.
     
  • Pilot 1 finishing in August 2021 

 

Technical Concept Overview

  • The goal is to predict wind events to adjust pitch and yaw to optimize energy capture and reduce gearbox stress.

 

  • Wind changes aloft are caused by and produce pressure changes due to conservation of energy.
     

  • Pressure changes near ground and well upstream and downstream from wind farms are related to wind changes at the wind farm. 
     

  • New highly sensitive, low-cost pressure sensors will be placed on short posts in an array around wind farms and extending out up to 500m.
     

  • Data from these sensors and wind farm anemometers will be collected on a wireless mesh network.
     

  • The AI model will be trained over time (some months) in supervised learning on pressure sensor array and wind data.
     

  • The edge AI model will compute real time inference on live pressure data to give wind change alerts to the owner’s SCADA. 
     

  • The 10-45 second alerts will allow proactive adjustments to pitch, yaw and other systems.
     

Atmospheric Science

Known factors

  • Pressure changes (Δp) are the drivers of air movement (wind).
  • The Bernoulli Principle drives boundary layer Δp from moving air.
  • Local and regional weather, thermals, etc. also drive wind and Δp.
  • Turbulence will also show up in high frequency Δp.
  • Δp can be measured very accurately and rapidly.

 

Expected but still to be demonstrated:

  • Area wide measurements of Δp near ground level correlates to wind vectors at the wind farm.
  • Correlated patterns repeat and therefore allow AI prediction.
  • Atmospheric events 60m or more above ground quickly produce Δp at the sensor array.
  • These effects are significant enough to measure:

             - Pressure drops downwind creating high p to low p movement.

             - Intersection of two moving air masses.

             - Partial heating from partial cloud cover.

  • Weather system movement (e.g. thunderstorm).

 

Our June 2021 field tests are starting to answer these questions.

Showing Windscape.ai price will be a fraction of the annual increased revenue.

Simple Value Proposition 

 

ASSUMPTIONS:

    Wind farm size: 50 MW

    Annual output*: 153,300 MWh/yr

           (Based on 35% CF - average for US fleet)

    Annual revenue*: $6,132,000

           (Based on $40 PPA - low range)

    Output improvement with windscape.ai: 1% 

           (Anticipate 2%-3%. Using lowest estimate here.)

    Annual energy production (AEP) only. Turbine life improvements not included.

 

RESULT:

    Increased wind farm annual revenue: $61,320

Team

Richard D. Ely, Ph.D., Inventor, Advisor

 

- Davis Wind, Founder

- Davis Hydro, Founder

 

- Massachusetts Institute of Technology

  • BS Geophysics
  • Graduate Research OR and EE

- UC Berkeley

  • MS Civil Engineering

- University of Rhode Island

  • MA Resource Economics

- University of Connecticut

  • MS Economics
  • Ph.D. Resource Economics

Eric Thompson, CEO

 

- Extensive leadership in wind and renewable energy

- Ensemble Energy (AI for wind farm O&M)

Chief Business Development Officer

- Nordic Windpower (Utility-scale WTG OEM)

  • Co-Founder 
  • VP Business Development 

- Strategy and Bizdev Consultant with multiple wind, solar and hydro technology companies

 

- Cornell University

  • BS Mechanical and Aerospace Engineering

Layne Williams, Chief Engineer

 

- 30+ years experience in Test, Network Design, System Development and Performance testing.

- Telkonet Inc. - Sr. System Test Engineer

- Uptime Network Inc. - Network Engineer

- Broadsoft Inc. - Test Engineer

- Implementing on Win10 / Win7, Linux RedHat Fedora Debian and Raspbian on RPi2/3/4/Arduino; working in Perl, C / C++, VBScript, Visual Basic, HTML, XML, Java and SQL scripting.  

Matthew A. Wright, Ph.D., AI/ML Lead

 

- Extensive controls, AI and traffic nodes modeling experience

- CV

 

- UC Berkeley

  • Postdoctoral Scholar: Berkeley AI Research (BAIR), Berkeley Deep Drive (BDD), Berkeley RISELab
  • PhD, Mechanical Engineering (controls engineering focus)

 

- UT Austin

  • BS, Mechanical Engineering
  • BA, Government

 

Nathan Sackheim, Analytics Developer

 

- Experienced in programming and data management with Python and PyTorch

 

- University of California, Santa Barbara

  • BS Physics
  • BS Mathematics

Nick Werhren, Full Stack Software Engineer

 

- Experienced with MVC web architecture

- Relational Database Expertise

 

- University Nevada Reno

  • BS Computer Science Engineering

Next Steps

Pilot 1 started in June with the helpful assistance of the UC Davis Atmospheric Science group at their test site.

  • 20 node array with pressure sensors in 300m radius area
  • Wifi base station and communications gateway
  • Tower anemometer data feed

AI analysis of data

  • Demonstrate Δp at the sensor array.
  • Write up of results

Initial fundraising

Phase 2 testing

  • Gen 2 sensor nodes with solar power
  • Edge AI capability
  • Multiple test sites
  • Live data analysis

Commercial Pilot

 

Contact

info@windscape.ai

 

Berkeley, CA