SSWSStd - Site-Specific Wind Speed + Standard Deviation Adjustment

Location: tact/adjustments/SSWSStd.py

Overview

The SSWSStd method extends the SSWS approach by adjusting both wind speed AND standard deviation independently before recalculating turbulence intensity. This provides two-parameter correction for comprehensive measurement adjustment.

Methodology

Algorithm Steps

Train Two Independent Regressions (on training data):

Reference_WS = m_ws * RSD_WS + c_ws
Reference_SD = m_sd * RSD_SD + c_sd

Apply Both Adjustments (to test data):

Adjusted_WS = m_ws * RSD_WS + c_ws
Adjusted_SD = m_sd * RSD_SD + c_sd

Recalculate Turbulence Intensity:

Adjusted_TI = Adjusted_SD / Adjusted_WS

Calculate Representative TI:

Adjusted_RepTI = Adjusted_TI + 1.28 * (Adjusted_SD / Adjusted_WS)

Key Characteristics

Dual adjustment: Corrects both WS and SD separately
Four parameters: Two slopes + two intercepts
Independent corrections: WS and SD adjusted separately, then combined
Higher complexity: More parameters than SSWS but potentially better accuracy
Error propagation: Still subject to error amplification in division

Usage

Basic Example

from tact import TACT

# Initialize
tact = TACT()

# Run SSWSStd adjustment
results = tact.adjust(
    data=your_data,
    method="sswsstd",
    parameters={
        "split": True,              # Enable train/test split
        "config_path": "config.json"  # Path to configuration
    }
)

Complete Workflow

from tact import TACT
from tact.utils.load_data import load_data
from tact.utils.setup_processors import setup_processors
from tact.validation import validate_dnv_rp0661
from tact.visualization import plot_dnv_validation

# Load and process data
data = load_data("data.csv")
binning_proc, ti_proc, stats_proc = setup_processors("config.json")
data = binning_proc.process(data)
data = ti_proc.process(data)

# Run SSWSStd adjustment
tact = TACT()
results = tact.adjust(
    data=data,
    method="sswsstd",
    parameters={"split": True, "config_path": "config.json"}
)

# Validate
validation = validate_dnv_rp0661(
    adjusted_data=results["adjusted_data"],
    reference_col="ref_ti",
    adjusted_col="adjTI_RSD_TI",
    wind_speed_col="ref_ws",
    bin_col="bins",
    criteria_type="LV"
)

# Visualize
plot_dnv_validation(
    validation_results=validation,
    adjusted_data=results["adjusted_data"],
    reference_col="ref_ti",
    adjusted_col="adjTI_RSD_TI",
    unadjusted_col="rsd_ti",
    wind_speed_col="ref_ws",
    bin_col="bins",
    method_name="SSWSStd",
    output_dir="output/plots"
)

Parameters

Required Parameters

Parameter	Type	Description
`config_path`	`str`	Path to configuration JSON file
`split`	`bool`	Whether to split data into train/test sets

Configuration Requirements

The config file must specify complete column mappings:

{
    "input_data_column_mapping": {
        "reference": {
            "wind_speed": "ref_ws",
            "wind_speed_std": "ref_sd",
            "turbulence_intensity": "ref_ti"
        },
        "rsd": {
            "primary": {
                "wind_speed": "rsd_ws",
                "wind_speed_std": "rsd_sd",
                "turbulence_intensity": "rsd_ti"
            }
        }
    }
}

Required Data Columns

The input data must contain:

Reference wind speed (e.g., ref_ws)
Reference wind speed standard deviation (e.g., ref_sd)
Reference turbulence intensity (e.g., ref_ti)
RSD wind speed (e.g., rsd_ws)
RSD wind speed standard deviation (e.g., rsd_sd)
RSD turbulence intensity (e.g., rsd_ti)
Train/test split indicator (e.g., split)
Wind speed bins (e.g., bins)

Output Format

Returned Dictionary

{
    "adjusted_data": pd.DataFrame,  # Data with adjusted columns
    "reg_results": pd.DataFrame,    # Regression statistics for both models
    "all_stats": pd.DataFrame       # Post-adjustment statistics
}

Adjusted Data Columns

The method adds these columns to your data:

Column	Description
`RSD_adjWS`	Adjusted RSD wind speed
`RSD_adjSD`	Adjusted RSD standard deviation
`adjTI_RSD_TI`	Adjusted turbulence intensity
`adjRepTI_RSD_RepTI`	Adjusted representative TI

Regression Results

Contains two rows (one for WS model, one for SD model):

Column	Description
`sensor`	Sensor identifier (WS or SD)
`height`	Measurement height
`adjustment`	Adjustment method name
`m`	Regression slope
`c`	Regression intercept
`rsquared`	R² value

Performance Characteristics

When SSWSStd Works Well

Both WS and SD have strong linear correlations with reference
Systematic bias in both measurements
High R² (>0.9) for both regression models
Consistent relationships across wind speed ranges

When SSWSStd May Struggle

Dual error propagation: Errors from both WS and SD adjustments combine
Low wind speeds: Small denominators amplify errors
Overfitting risk: Four parameters may overfit to training data
Non-linear relationships: Linear models can’t capture complex patterns
Poor correlation: Low R² in either model degrades performance

Performance on Example Dataset

Based on DNV RP-0661 LV criteria validation:

Metric	Value	Target	Pass
MRBE	+37.73%	≤5%	❌
RRMSE	121.61%	≤15%	❌
N observations	3,137	-	-

Ranking: 2nd best MRBE, 3rd best RRMSE (out of 4 methods) Analysis: Better than SSWS but not as good as SS-SF. The dual adjustment adds complexity without improving performance enough to justify it. Error propagation through division still dominates.

Comparison with Other Methods

Performance Comparison

Method	MRBE	RRMSE	Complexity	Rank
SS-SF	37.80%	111.13%	Low	🥇 1st
SSWSStd	37.73%	121.61%	High	🥈 2nd
SSWS	92.06%	189.39%	Medium	🥉 3rd
Baseline	90.39%	183.04%	None	-

Feature Comparison

Aspect	SSWSStd	SS-SF	SSWS
Parameters	4 (m_ws, c_ws, m_sd, c_sd)	2 (m, c)	2 (m, c)
Adjusts	WS + SD	TI directly	WS only
Error Propagation	Yes (dual)	No	Yes
Training Complexity	Two models	One model	One model
Best For	Complex bias in both WS and SD	General purpose	WS-only bias

Implementation Details

Class Definition

from tact.core.base import AdjustmentMethod
from tact.core.registry import AdjustmentRegistry

@AdjustmentRegistry.register("sswsstd")
class SSWSStd(AdjustmentMethod):
    """Site-Specific Wind Speed + Standard Deviation adjustment."""

    def required_model_parameters(self) -> Dict[str, type]:
        return {"config_path": str, "split": bool}

    def required_data_columns(self) -> list:
        return [
            "reference.wind_speed",
            "reference.wind_speed_std",
            "reference.turbulence_intensity",
            "rsd.height_1.wind_speed",
            "rsd.height_1.wind_speed_std",
            "rsd.height_1.turbulence_intensity"
        ]

    def adjust(self, data: pd.DataFrame, parameters: Dict[str, Any]):
        # Implementation...

Mathematical Formulation

The method solves two independent least-squares problems: Wind Speed Model:

minimize: Σ(Reference_WS - (m_ws * RSD_WS + c_ws))²

Standard Deviation Model:

minimize: Σ(Reference_SD - (m_sd * RSD_SD + c_sd))²

Then combines them:

TI = Adjusted_SD / Adjusted_WS

Troubleshooting

High MRBE/RRMSE Despite Good R² Values

Cause: Error propagation when dividing adjusted SD by adjusted WS Solution:

Try SS-SF method (adjusts TI directly, no division)
Consider filtering low wind speed data where division errors are largest
Check for systematic bias patterns by wind speed bin

Large Differences Between Training and Test Performance

Cause: Overfitting with 4 parameters, or non-representative split Solution:

# Verify train/test split
train = data[data['split'] == 1]
test = data[data['split'] == 0]

print(f"Train size: {len(train)}, Test size: {len(test)}")
print(f"Train WS range: {train['ref_ws'].min():.1f}-{train['ref_ws'].max():.1f} m/s")
print(f"Test WS range: {test['ref_ws'].min():.1f}-{test['ref_ws'].max():.1f} m/s")

Negative Adjusted Values

Cause: Large negative intercepts with low WS or SD values Solution:

# Check for negative values
adjusted = results["adjusted_data"]
neg_ws = (adjusted['RSD_adjWS'] < 0).sum()
neg_sd = (adjusted['RSD_adjSD'] < 0).sum()

if neg_ws > 0 or neg_sd > 0:
    print(f"⚠️  Warning: {neg_ws} negative WS, {neg_sd} negative SD")
    print("Consider filtering low-value training data")

When to Use SSWSStd

Use SSWSStd When:

Both WS and SD show systematic bias vs reference
You have large dataset (>2000 points for training)
R² > 0.9 for both WS and SD models
You need independent control over WS and SD corrections

❌ Don’t Use SSWSStd When:

SS-SF performs better (check comparison results)
You have limited data (less than 500 points)
Either WS or SD has poor correlation (R² less than 0.8)
Error propagation is a concern (low wind speeds common)

References

DNV GL: DNV-RP-0661 - Remote Sensing Measurement Verification
IEC 61400-12-1 - Wind turbine power performance testing

Source Code

Full implementation: tact/adjustments/SSWSStd.py

API Reference

Sswsstd

SSWSStd - Site-Specific Wind Speed + Standard Deviation Adjustment

Overview

Methodology

Algorithm Steps

Key Characteristics

Usage

Basic Example

Complete Workflow

Parameters

Required Parameters

Configuration Requirements

Required Data Columns

Output Format

Returned Dictionary

Adjusted Data Columns

Regression Results

Performance Characteristics

When SSWSStd Works Well

When SSWSStd May Struggle

Performance on Example Dataset

Comparison with Other Methods

Performance Comparison

Feature Comparison

Implementation Details

Class Definition

Mathematical Formulation

Troubleshooting

High MRBE/RRMSE Despite Good R² Values

Large Differences Between Training and Test Performance

Negative Adjusted Values

When to Use SSWSStd

Use SSWSStd When:

❌ Don’t Use SSWSStd When:

See Also

References

Source Code

​SSWSStd - Site-Specific Wind Speed + Standard Deviation Adjustment

​Overview

​Methodology

​Algorithm Steps

​Key Characteristics

​Usage

​Basic Example

​Complete Workflow

​Parameters

​Required Parameters

​Configuration Requirements

​Required Data Columns

​Output Format

​Returned Dictionary

​Adjusted Data Columns

​Regression Results

​Performance Characteristics

​When SSWSStd Works Well

​When SSWSStd May Struggle

​Performance on Example Dataset

​Comparison with Other Methods

​Performance Comparison

​Feature Comparison

​Implementation Details

​Class Definition

​Mathematical Formulation

​Troubleshooting

​High MRBE/RRMSE Despite Good R² Values

​Large Differences Between Training and Test Performance

​Negative Adjusted Values

​When to Use SSWSStd

​Use SSWSStd When:

​❌ Don’t Use SSWSStd When:

​See Also

​References

​Source Code

SSWSStd - Site-Specific Wind Speed + Standard Deviation Adjustment

Overview

Methodology

Algorithm Steps

Key Characteristics

Usage

Basic Example

Complete Workflow

Parameters

Required Parameters

Configuration Requirements

Required Data Columns

Output Format

Returned Dictionary

Adjusted Data Columns

Regression Results

Performance Characteristics

When SSWSStd Works Well

When SSWSStd May Struggle

Performance on Example Dataset

Comparison with Other Methods

Performance Comparison

Feature Comparison

Implementation Details

Class Definition

Mathematical Formulation

Troubleshooting

High MRBE/RRMSE Despite Good R² Values

Large Differences Between Training and Test Performance

Negative Adjusted Values

When to Use SSWSStd

Use SSWSStd When:

❌ Don’t Use SSWSStd When:

See Also

References

Source Code