IEA Wind Recommended Practice for the Implementation of Renewable Energy Forecasting Solutions

Front Cover
IEA Wind Recommended Practice for the Implementation of Renewable Energy Forecasting Solutions
Copyright
Contents
List of figures
List of tables
Biography
Dr. Corinna Möhrlen
Dr. John W. Zack
Dr. Gregor Giebel
Preface
About the IEA Wind TCP and Task 36 and 51
Part 1 Forecast solution selection process
1 Forecast solution selection process
1.1 Before you start reading
1.2 Background and introduction
1.3 Objectives
1.4 Definitions
2 Initial considerations
2.1 Tackling the task of engaging a forecaster for the first time
2.2 Purpose and requirements of a forecasting solution
2.3 Adding uncertainty forecasts to forecasting solutions
2.4 Information table for specific topic targets
3 Decision support tool
3.1 Initial forecast system planning
3.2 IT infrastructure considerations
3.2.1 IT requirements for single versus multiple forecast vendors
3.2.2 IT requirements for deterministic versus probabilistic forecasts
3.3 Establishment of a requirement list
3.3.1 Requirement list
3.4 Short-term solution
3.5 Long-term solution
3.6 Going forward with an established IT system
3.7 Complexity level of the existing IT solution
3.8 Selection of a new vendor versus benchmarking existing vendor
3.9 RFP evaluation criteria for a forecast solution
3.9.1 Forecast solution type
3.9.1.1 Single versus multiple forecast providers
3.9.1.2 Deterministic versus probabilistic
3.9.1.3 Forecast horizons
3.9.2 Vendor capabilities
3.9.2.1 Experience and reliability
3.9.2.2 Ability to maintain state-of-the-art performance
3.9.2.3 Performance incentive schemes
3.9.3 Evaluation of services
3.9.3.1 Price versus value and quality
3.9.3.2 Forecast performance
3.9.3.3 Solution characteristics
3.9.3.4 Support structure
3.9.3.5 Redundancy structure
3.9.3.6 Escalation structure
3.10 Forecast methodology selection for use of probabilistic forecasts
3.10.1 Definitions of uncertainty
3.10.2 Uncertainty forecasting methods
3.10.3 Training tools for ensemble forecasting
3.10.4 Applications of uncertainty forecasts in the energy industry
3.10.5 Visualization of forecast uncertainty
4 Data communication
4.1 Terminology
4.2 Data description
4.2.1 LEVEL 1 – data description
4.3 Data format and exchange
4.3.1 LEVEL 1 data format and exchange
4.3.2 LEVEL 2 – data format and exchange
4.4 Sample formatted template files and schemas
5 Concluding remarks
Part 2 Designing and executing forecasting benchmarks and trials
6 Designing and executing benchmarks and trials
6.1 Before you start reading
6.2 Background and introduction
6.3 Definitions
6.3.1 Renewable energy forecast benchmark
6.3.2 Renewable energy forecast trial
6.4 Objectives
7 Initial considerations
7.1 Deciding whether to conduct a trial or benchmark
7.2 Benefits of trials and benchmarks
7.3 Limitations with trials and benchmarks
7.4 Time lines and forecast periods in a trial or benchmark
7.5 1-Page ``cheat sheet'' checklist
8 Conducting a benchmark or trial
8.1 Phase 1: preparation
8.1.1 Key considerations in the preparation phase
8.1.2 Metadata gathering in the preparation phase
8.1.3 Historical data gathering in the preparation phase
8.1.4 IT/data considerations in the preparation phase
8.1.5 Communication in the preparation phase
8.1.6 Test run in the preparation phase
8.2 Phase 2: During benchmark/trial
8.2.1 Communication during the b/t
8.2.2 Forecast validation and reporting during the b/t
8.3 Phase 3: Post trial or benchmark
8.3.1 Communication at the end of the b/t
8.3.2 Forecast validation & reporting after the b/t
9 Considerations for probabilistic benchmarks and trials
9.1 Preparation phase challenges for probabilistic b/t
9.2 Evaluation challenges for probabilistic b/t
10 Best practice recommendations for benchmarks/trials
10.1 Best practice for b/t
10.2 Pitfalls to avoid
Part 3 Forecast solution evaluation
11 Forecast solution evaluation
11.1 Before you start reading
11.2 Background and introduction
12 Overview of evaluation uncertainty
12.1 Representativeness
12.1.1 Size and composition of the evaluation sample
12.1.2 Data quality
12.1.3 Forecast submission control
12.1.4 Process information dissemination
12.2 Significance
12.2.1 Quantification of uncertainty
12.2.1.1 Method 1: repeating the evaluation task
12.2.1.2 Method 2: bootstrap resampling
12.3 Relevance
13 Measurement data processing and control
13.1 Uncertainty of instrumentation signals and measurements
13.2 Measurement data reporting and collection
13.2.1 Non-weather related production reductions
13.2.2 Aggregation of measurement data in time and space
13.3 Measurement data processing and archiving
13.4 Quality assurance and quality control
14 Assessment of forecast performance
14.1 Forecast attributes at metric selection
14.1.1 ``Typical'' error metrics
14.1.2 Outlier/extreme error
14.1.3 Empirical error distribution
14.1.4 Binary or multi-criteria events
14.2 Prediction intervals and predictive distributions
14.3 Probabilistic forecast assessment methods
14.3.1 Brier scores
14.3.2 Ranked probability (skill) score (RP(S)s)
14.3.2.1 The continuous ranked probability skill and energy score
14.3.2.2 Logarithmic and variogram scoring rules
14.3.3 Reliability measures
14.3.3.1 Rank histogram
14.3.3.2 Reliability (calibration) diagram
14.3.4 Event discrimination ability: relative operating characteristic (ROC)
14.3.5 Uncertainty in forecasts: Rény entropy
14.4 Metric-based forecast optimization
14.5 Post-processing of ensemble forecasts
15 Best practice recommendations for forecast evaluation
15.1 Developing an evaluation framework
15.1.1 Scoring rules for comparison of forecast types
15.1.2 Forecast and forecast error analysis
15.1.3 Choice of deterministic verification methods
15.1.3.1 Dichotomous event evaluation
15.1.3.2 Analyzing forecast error spread with box and wiskers plots
15.1.3.3 Visualizing the error frequency distribution with histograms
15.1.4 Specific probabilistic forecast verification
15.1.4.1 Choice of application for benchmarking probabilistic forecasts
15.1.5 Establishing a cost function or evaluation matrix
15.1.5.1 Evaluation matrix
15.2 Operational forecast value maximization
15.2.1 Performance monitoring
15.2.1.1 Importance of performance monitoring for different time periods
15.2.2 Forecast diagnostics and continuous improvement
15.2.3 Maximization of forecast value
15.2.4 Maintaining state-of-the-art performance
15.2.4.1 Significance test for new developments
15.2.5 Incentivization
15.3 Evaluation of benchmarks and trials
15.3.1 Applying the principles of representativeness, significance, and relevance
15.3.2 Evaluation preparation in the execution phase
15.3.3 Performance analysis in the evaluation phase
15.3.4 Evaluation examples from a benchmark
15.4 Use cases
15.4.1 Energy trading and balancing
15.4.1.1 Forecast error cost functions
15.4.2 General ramping forecasts
15.4.2.1 Amplitude versus phase
15.4.2.2 Costs of false alarms
15.4.3 Evaluation of probabilistic ramp forecasts for reserve allocation
15.4.3.1 Definition of error conditions for the forecast
Part 4 Meteorological and power data requirements for real-time forecasting applications
16 Meteorological and power data requirements for real-time forecasting applications
16.1 Before you start reading
16.2 Background and introduction
16.3 Structure and recommended use
17 Use and application of real-time meteorological measurements
17.1 Application-specific requirements
17.1.1 Application-specific requirements for meteorological data
17.1.2 Applications in system operation, balancing and trading
17.1.3 Applications in wind turbine and wind farm operation
17.1.4 Solar/PV plant operation
17.2 Available and applicable standards for real-time meteorological and power measurements
17.2.1 Standards and guidelines for wind measurements
17.2.2 Standards and guidelines for solar measurements
17.3 Standards and guidelines for general meteorological measurements
17.4 Data communication
18 Meteorological instrumentation for real-time operation
18.1 Instrumentation for wind projects
18.1.1 Cup anemometers
18.1.2 Sonic and ultra-sonic anemometers
18.1.3 Remote sensing devices
18.1.4 Met mast sensor deployment
18.1.5 Nacelle sensor deployment
18.2 Instrumentation for solar projects
18.2.1 Point measurements
18.2.2 All sky imagers
18.2.3 Satellite data
19 Power measurements for real-time operation
19.1 Live power and related measurements
19.2 Measurement systems
19.2.1 Connection-point meters
19.2.2 Wind power SCADA systems
19.2.3 Solar power SCADA systems
19.3 Power available signals
19.3.1 Embedded wind and solar ``behind the meter''
19.4 Live power data in forecasting
19.4.1 Specifics for producers of forecasts
19.4.2 Specifics for consumers/users of forecasts
19.5 Summary of best practices
20 Measurement setup and calibration
20.1 Selection of instrumentation
20.1.1 Selection of instrumentation for wind projects
20.1.2 Selection of instrumentation for solar power plants
20.1.3 Measurement characteristics of different technologies
20.1.3.1 Measurement characteristics of LIDARs
20.1.3.2 Lightning effects on instrumentation
20.1.3.3 Measurement characteristics of SODARs
20.2 Location of measurements
20.2.1 Location of representative measurements for wind projects
20.2.2 Location of representative measurements for solar projects
20.3 Maintenance and inspection schedules
20.3.1 Maintenance of radiometers
21 Assessment of instrumentation performance
21.1 Measurement data processing
21.2 Uncertainty expression in measurements
21.3 Known issues of uncertainty in specific instrumentation
21.3.1 Effects of uncertainty in nacelle wind speed measurements and mitigation methods
21.3.1.1 Wind speed reduction in the induction zone
21.3.1.2 Wake effects from rotating blades
21.3.1.3 Yaw misalignment of wind turbine for scanning LIDARs
21.3.2 Application of nacelle wind speeds in real-time NWP data assimilation
21.3.3 Known uncertainty in radiation measurements
21.4 General data quality control and quality assurance (QCQA)
21.5 Historic quality control (QC)
21.5.1 QC for wind forecasting applications
21.5.1.1 Specific control procedures
21.5.1.2 Practical methodology for quality control of measurement for wind applications
21.5.1.3 Statistical tests and metrics for the QC process
21.5.2 QC for solar applications
21.6 Real-time quality control (QC)
21.6.1 Data screening in real-time wind and solar forecast applications
21.6.2 Data sampling thresholds in real-time wind and solar forecast applications
21.6.3 Real-time QC for wind and solar applications
21.6.3.1 Data screening
21.6.4 Solar forecasting specific real-time QC
22 Best practice recommendations
22.1 Definitions
22.2 Instrumentation
22.3 Recommendations for real-time measurements by application type
22.4 Recommendations for real-time measurements for power grid and utility-scale operation
22.4.1 Recommendations on quality requirements
22.4.1.1 Requirements for wind forecasting applications according to environment
22.4.1.2 Wind measurement alternatives to met masts
22.4.1.3 Recommendations for solar forecasting applications
22.4.1.4 Recommendations for power measurements for real-time wind and solar forecasting
22.4.2 Accuracy and resolution recommendations
22.4.3 Validation and verification recommendations
22.4.3.1 Practical methodology for historic quality control of meteorological measurements
22.4.3.2 Data screening in real-time environments
22.5 Recommendations for real-time measurements for power plant operation and monitoring
22.5.1 Data quality recommendations
22.5.1.1 Requirement recommendations for wind farms
22.5.1.2 Requirement recommendations for wind farms using remote sensing
22.5.1.3 Requirement recommendations for solar plants
22.5.2 Validation and verification recommendations
22.5.2.1 Statistical test and metric recommendations for the QC process
22.5.2.2 Solar specific validation recommendations
22.5.2.3 Performance control recommendations for hardware and manufacturer production guarantees
22.6 Recommendations for real-time measurements for power trading in electricity markets
22.6.1 Trading strategies with real-time measurements
22.6.2 Quality recommendations
22.6.3 Accuracy and resolution requirement recommendations
23 End note
A Clarification questions for forecast solutions
Ask questions to the vendors
Typical questions for Part 1
Typical questions for Part 2
B Typical RFI questions prior to or in an RFP
C Application examples for use of probabilistic uncertainty forecasts
C.1 Example of the graphical visualization of an operational dynamic reserve prediction system at a system operator
C.2 High-speed shut down warning system
D Metadata checklist
E Sample forecast file structures
E.1 XSD template example for forecasts and SCADA
E.2 XSD SCADA template for exchange of real-time measurements
F Standard statistical metrics
F.1 BIAS
F.2 MAE – mean absolute error
F.3 RMSE – root mean squared error
F.4 Correlation
F.5 Standard deviation
F.6 What makes a forecast ``good''?
G Validation and verification code examples
G.1 IEA wind task 36 and task 51 specific V&V code examples
G.2 Code examples from related projects with relevance to recommendations
H Examples of system operator met measurement requirements
H.1 Examples of requirements in different jurisdictions
H.2 Met measurement requirement example from California independent system operator in USA
H.3 Met measurement requirement example from Irish system operator EIRGRID group
H.4 Met measurement requirement example from Alberta electric system operator in Canada
Bibliography
Nomenclature
Index
Back Cover