How to reduce risk




Presented by:
Matthew Hendrickson, Sr. Director Of Assessment
3TIER is Focused on Understanding the Fuel

» Founded in 1999
» Headquarters in Seattle, WA
» Offices in Panama, India, and
  Australia
» Focused on renewable energy
  information services
    39,000 MW wind energy
     forecasting
    7,400 MW hydropower
     forecasting
    Extensive international
     wind & solar resource
     assessment
Pre-construction services
I.          Spatial mapping
II.         Climate variability analysis
III.        Comprehensive net report
       A.     Spatial
       B.     Climate
       C.     Wake
       D.     Other losses
       E.     Uncertainty
What is Numerical Weather Prediction?


 INPUTS           WRF ANALYSIS      OUTPUTS
   Global
  Weather
  Archive
1960-present
                                 Understanding of
                                 wind characteristics
     High
  Resolution                     Long-term variability
 Terrain, Soil,                  assessments
and Vegetation
     Data                        (up to 50 years)

                                 Spatial wind maps
  OPTIONAL:
    Onsite
 Observations
The Value of Higher Resolution Wind Mapping




  5km resolution                1.5km resolution               500m resolution


8.0
      Look at the two locations marked on each map…
                                Sweetwater       Nearby Mountain
      At 5km resolution =           ~7.1m/s        ~7.5m/s
      At 1.5km resolution =         ~6.9m/s        ~7.7m/s
7.0
      At 500m resolution =          ~6.8m/s        ~8.0m/s



6.0
The Value of High Resolution Wind Mapping




  5km resolution       1.5km resolution           500m resolution


8.0

                                          . . . and understand the
                                          spatial variability of the wind
7.0                90m resolution         resource, better than with
                                          simple statistical
                   ~9.0 m/s at mountain   extrapolations and/or
6.0                                       interpolations of on-site
                                          observations
Avoid Wind Holes!
Options for Understanding Long-term Variability
Using Nearby 3rd Party Station Data



       Measure-Correlate-Predict
                (MCP)
Traditional MCP
Measure-Correlate-Predict


               Reference ✪
               Site

                                             ws=m*wsr+b




                             Uses a statistical relationship
                             between on-site obs data and a
                             longer ‘reference’ site to
                             understand the variability of the
                             wind resource and determine a
                             long-term adjustment
Traditional MCP
Measure-Correlate-Predict

 » Reference site needs to be a consistent, long-term time
   series located within a similar flow regime as the project site
 » For robust results, MCP requires high correlation between
   reference site and on-site, project data




 » What to do if suitable long-term reference data are not
   available?
Using Synthetic Reference Data



 Numerical Weather Prediction Models
Synthetic Reference Data – Using NWP Models

» Unlike MCP analysis, Numerical Weather Prediction (NWP)
  models do not require off-site reference data.
» Over 50 years of historic wind resource data at hourly
  resolution can be generated utilizing NWP models
» Synthetic reference data are provided at the project-site
Numerical Weather Prediction Framework

     Input                                Output

Global Weather                          Understanding
   Archive                                 of Wind
1948-present                            Characteristics
                        Numerical
                         Weather
High Resolution                          Long-Term
                     Prediction Model
Terrain, Soil, and                       Variability
Vegetation Data           (NWP)         Assessments


   On-Site                                Spatial
 Observations                            Wind Maps
NWP Output – Annual-mean Variability




            NWP wind speed   Observed wind speed
            Long-term mean   Operational wind speed
NWP Output – Monthly-mean Variability
Capacity Factor




                  Based on single year of observations, compared to 40 year analysis
Validation of 3TIER NWP Time Series



      Comparison Against Reanalysis
Skill of Reanalysis Data

          287 QC’d Tall Towers
          Average R=0.53
          Median R=0.55
          STD R=0.19
Skill of 3TIER Data

          287 QC’d Tall Towers
          Average R=0.70
          Median R=0.71
          STD R=0.08
Skill of 3TIER Data - Independent Validation
                    Wind energy index comaprison                                                            Wind energy index comaprison
          200                                                                                    200
                                                                                                           y = 0,9692x + 2,6531
          180                                                                                    180            R² = 0,8772

          160                                                                                    160
                    y = 0,5507x + 44,931
                         R² = 0,6706
          140                                                                                    140

          120                                                                                    120




                                                                                     3Tier 20m
   NCAR




          100                                                                                    100

           80                                                                                     80

           60                                                                                     60


           40                                                                                     40


           20                                                                                     20


            0                                                                                      0

                0     20     40      60    80   100    120   140   160   180   200                     0      20     40      60   80   100    120   140   160   180   200

                                           ElZayt NW 24,5m                                                                        ElZayt NW 24,5m




    "Our preliminary findings show that 3TIER wind data can
   significantly reduce long-term correction error compared to
   using NCAR/NCEP data, which is often the only option due to a
   lack of reliable long-term ground measurements.

   Per Nielsen - EMD Manager
Incorporating On-site Observational Data
                  with
  Numerical Weather Prediction Models

           Model Output Statistics
                   (MOS)
Incorporating Observational Data

At any particular location the best way to
determine the wind resource is through direct
measurement.
3TIER incorporates observational data into
wind assessments whenever suitable on-site
data are available for
    • Validation
    • Statistical correction
Incorporating Observational Data - MOS


• MOS (Model Output Statistics) is a statistical technique to
  remove bias & adjust the variance of NWP model data to
  better match the on-site observed data
• NWP models simulate the full structure and time evolution of
  the atmosphere. MOS relates the observed wind speed to
  the leading NWP predictors to improve the quality of the
  long-term estimate of the wind resource.
• The output of the MOS algorithm is a multi-linear equation
  that is applied to all times of the analysis on an hourly basis
  (windspeed_97m * 1.25) + (windspeed_200m * 0.51) + (u_200m * 0.07) +
    (temperature_0m * -0.06) + 18.93
MOS-Corrected Output – Monthly-mean Variability
Validation of MOS-Corrected Time Series



     Influence of observational record length
Skill at Monthly-mean Timescale


        Raw Model Data                 MOS-Corrected Model Data




 Analysis performed with Horizon Wind Energy utilizing 299 tall towers

 MOS-corrected standard deviation of error for an individual month = 7.8%
Skill at Annual-mean Timescale


        Raw Model Data                 MOS-Corrected Model Data




 Analysis performed with Horizon Wind Energy utilizing 299 tall towers

 MOS-corrected standard deviation of error for an individual year = 3.4%
MOS skill utilizing short observational records




  A single month of observational data helps to remove bias

  MOS-corrected errors decrease throughout first year of observed record
MOS skill utilizing short observational records




  MOS-corrected errors decrease throughout first year of observed record
Comparing Skill of
3TIER MOS & MCP


 Results based on 23 met
 towers each with 5 years
 of obs data


 Obs data and MCP
 analysis provided by
 Horizon Wind Energy




                            3TIER MOS
                            MCP
Wake Modeling

• 3TIER’s super computing
  capabilities allow unique
  ability to model wakes in
  time series across all
  climatic conditions.

• Classical engineering
  solutions require climatic
  conditions to be condensed
  into distributions, disguising
  important features like
  performance in atmospheric
  stable conditions
Other losses
» Turbulence
» Shear
» Inflow angle
» Electrical system
» Availability
» Turbine performance
» Environmental
» Blade degradation
» Icing
» Wind sector management
» High speed start/stop hysteresis
Comprehensive Assessment

• 3TIER’s most complete solution, provides finance quality
  energy assessment
• Project-wide, net energy assessment based on the last 40+
  years of MOS-corrected NWP model data
• Adds a site-visit, quality control of obs data, full uncertainty
  analysis, and gross-to-net analysis to a FullView Project
  Resource Assessment
• Uses 3TIER’s proprietary time-varying wake modeling
  analysis to understand diurnal and seasonal variability of
  wakes, wind speed deficits, and turbulence intensity
Comprehensive Assessment




  Full, time series simulation at every turbine across historic record
Comprehensive Assessment




Multi-staged observation QC process
Comprehensive Assessment
Comprehensive Assessment
 Comprehensive uncertainty analysis highlights risk associated
 with measurements, shear, spatial modeling, temporal
 modeling, generation, wake modeling, etc…
Other Services

   • Power Performance Testing – per IEC 61400-
     12-1 standards
   • Operational reforecast services – reassess the
     long term production of a plant after operational
     data is available
   • Operational forensic services – root cause
     analysis to “deep dive” into SCADA data and
     attempt to explain variations on production
     against expectations.
   • Etc… 3TIER’s Advanced Applications group
     positioned to tackle any challenges in need of
     scientific solution.

I Workshop Wind GlobalGeo e 3TIER - Matt

  • 1.
    How to reducerisk Presented by: Matthew Hendrickson, Sr. Director Of Assessment
  • 2.
    3TIER is Focusedon Understanding the Fuel » Founded in 1999 » Headquarters in Seattle, WA » Offices in Panama, India, and Australia » Focused on renewable energy information services  39,000 MW wind energy forecasting  7,400 MW hydropower forecasting  Extensive international wind & solar resource assessment
  • 3.
    Pre-construction services I. Spatial mapping II. Climate variability analysis III. Comprehensive net report A. Spatial B. Climate C. Wake D. Other losses E. Uncertainty
  • 4.
    What is NumericalWeather Prediction? INPUTS WRF ANALYSIS OUTPUTS Global Weather Archive 1960-present Understanding of wind characteristics High Resolution Long-term variability Terrain, Soil, assessments and Vegetation Data (up to 50 years) Spatial wind maps OPTIONAL: Onsite Observations
  • 5.
    The Value ofHigher Resolution Wind Mapping 5km resolution 1.5km resolution 500m resolution 8.0 Look at the two locations marked on each map… Sweetwater Nearby Mountain At 5km resolution = ~7.1m/s ~7.5m/s At 1.5km resolution = ~6.9m/s ~7.7m/s 7.0 At 500m resolution = ~6.8m/s ~8.0m/s 6.0
  • 6.
    The Value ofHigh Resolution Wind Mapping 5km resolution 1.5km resolution 500m resolution 8.0 . . . and understand the spatial variability of the wind 7.0 90m resolution resource, better than with simple statistical ~9.0 m/s at mountain extrapolations and/or 6.0 interpolations of on-site observations
  • 7.
  • 8.
    Options for UnderstandingLong-term Variability
  • 9.
    Using Nearby 3rdParty Station Data Measure-Correlate-Predict (MCP)
  • 10.
    Traditional MCP Measure-Correlate-Predict Reference ✪ Site ws=m*wsr+b Uses a statistical relationship between on-site obs data and a longer ‘reference’ site to understand the variability of the wind resource and determine a long-term adjustment
  • 11.
    Traditional MCP Measure-Correlate-Predict »Reference site needs to be a consistent, long-term time series located within a similar flow regime as the project site » For robust results, MCP requires high correlation between reference site and on-site, project data » What to do if suitable long-term reference data are not available?
  • 12.
    Using Synthetic ReferenceData Numerical Weather Prediction Models
  • 13.
    Synthetic Reference Data– Using NWP Models » Unlike MCP analysis, Numerical Weather Prediction (NWP) models do not require off-site reference data. » Over 50 years of historic wind resource data at hourly resolution can be generated utilizing NWP models » Synthetic reference data are provided at the project-site
  • 14.
    Numerical Weather PredictionFramework Input Output Global Weather Understanding Archive of Wind 1948-present Characteristics Numerical Weather High Resolution Long-Term Prediction Model Terrain, Soil, and Variability Vegetation Data (NWP) Assessments On-Site Spatial Observations Wind Maps
  • 15.
    NWP Output –Annual-mean Variability NWP wind speed Observed wind speed Long-term mean Operational wind speed
  • 16.
    NWP Output –Monthly-mean Variability Capacity Factor Based on single year of observations, compared to 40 year analysis
  • 17.
    Validation of 3TIERNWP Time Series Comparison Against Reanalysis
  • 18.
    Skill of ReanalysisData 287 QC’d Tall Towers Average R=0.53 Median R=0.55 STD R=0.19
  • 19.
    Skill of 3TIERData 287 QC’d Tall Towers Average R=0.70 Median R=0.71 STD R=0.08
  • 20.
    Skill of 3TIERData - Independent Validation Wind energy index comaprison Wind energy index comaprison 200 200 y = 0,9692x + 2,6531 180 180 R² = 0,8772 160 160 y = 0,5507x + 44,931 R² = 0,6706 140 140 120 120 3Tier 20m NCAR 100 100 80 80 60 60 40 40 20 20 0 0 0 20 40 60 80 100 120 140 160 180 200 0 20 40 60 80 100 120 140 160 180 200 ElZayt NW 24,5m ElZayt NW 24,5m "Our preliminary findings show that 3TIER wind data can significantly reduce long-term correction error compared to using NCAR/NCEP data, which is often the only option due to a lack of reliable long-term ground measurements. Per Nielsen - EMD Manager
  • 21.
    Incorporating On-site ObservationalData with Numerical Weather Prediction Models Model Output Statistics (MOS)
  • 22.
    Incorporating Observational Data Atany particular location the best way to determine the wind resource is through direct measurement. 3TIER incorporates observational data into wind assessments whenever suitable on-site data are available for • Validation • Statistical correction
  • 23.
    Incorporating Observational Data- MOS • MOS (Model Output Statistics) is a statistical technique to remove bias & adjust the variance of NWP model data to better match the on-site observed data • NWP models simulate the full structure and time evolution of the atmosphere. MOS relates the observed wind speed to the leading NWP predictors to improve the quality of the long-term estimate of the wind resource. • The output of the MOS algorithm is a multi-linear equation that is applied to all times of the analysis on an hourly basis (windspeed_97m * 1.25) + (windspeed_200m * 0.51) + (u_200m * 0.07) + (temperature_0m * -0.06) + 18.93
  • 24.
    MOS-Corrected Output –Monthly-mean Variability
  • 25.
    Validation of MOS-CorrectedTime Series Influence of observational record length
  • 26.
    Skill at Monthly-meanTimescale Raw Model Data MOS-Corrected Model Data Analysis performed with Horizon Wind Energy utilizing 299 tall towers MOS-corrected standard deviation of error for an individual month = 7.8%
  • 27.
    Skill at Annual-meanTimescale Raw Model Data MOS-Corrected Model Data Analysis performed with Horizon Wind Energy utilizing 299 tall towers MOS-corrected standard deviation of error for an individual year = 3.4%
  • 28.
    MOS skill utilizingshort observational records A single month of observational data helps to remove bias MOS-corrected errors decrease throughout first year of observed record
  • 29.
    MOS skill utilizingshort observational records MOS-corrected errors decrease throughout first year of observed record
  • 30.
    Comparing Skill of 3TIERMOS & MCP Results based on 23 met towers each with 5 years of obs data Obs data and MCP analysis provided by Horizon Wind Energy 3TIER MOS MCP
  • 31.
    Wake Modeling • 3TIER’ssuper computing capabilities allow unique ability to model wakes in time series across all climatic conditions. • Classical engineering solutions require climatic conditions to be condensed into distributions, disguising important features like performance in atmospheric stable conditions
  • 32.
    Other losses » Turbulence »Shear » Inflow angle » Electrical system » Availability » Turbine performance » Environmental » Blade degradation » Icing » Wind sector management » High speed start/stop hysteresis
  • 33.
    Comprehensive Assessment • 3TIER’smost complete solution, provides finance quality energy assessment • Project-wide, net energy assessment based on the last 40+ years of MOS-corrected NWP model data • Adds a site-visit, quality control of obs data, full uncertainty analysis, and gross-to-net analysis to a FullView Project Resource Assessment • Uses 3TIER’s proprietary time-varying wake modeling analysis to understand diurnal and seasonal variability of wakes, wind speed deficits, and turbulence intensity
  • 34.
    Comprehensive Assessment Full, time series simulation at every turbine across historic record
  • 35.
  • 36.
  • 37.
    Comprehensive Assessment Comprehensiveuncertainty analysis highlights risk associated with measurements, shear, spatial modeling, temporal modeling, generation, wake modeling, etc…
  • 38.
    Other Services • Power Performance Testing – per IEC 61400- 12-1 standards • Operational reforecast services – reassess the long term production of a plant after operational data is available • Operational forensic services – root cause analysis to “deep dive” into SCADA data and attempt to explain variations on production against expectations. • Etc… 3TIER’s Advanced Applications group positioned to tackle any challenges in need of scientific solution.