Performance+ is Bazefield's native advanced-analytics product for utility-scale wind. It turns raw turbine SCADA and meteorological data into a complete, auditable picture of how a site is performing, why, and what it is costing - production, losses attributed to their root cause, turbine health, degradation, and prioritised O&M actions.

This document is the entry point to the Performance+ wind documentation. It summarises what the product does, what makes it different, and how the underlying pipeline fits together, then links out to the detailed technical notes for each capability.

What Performance+ delivers

From the same input data and the same status-detection layer, Performance+ produces:

  • A self-learned expected-power reference per turbine - the yardstick every loss is measured against.
  • A full loss breakdown that attributes every lost kWh (and its revenue) to a specific root cause - downtime, derate, curtailment, icing, yaw misalignment and wake loss.
  • Turbine health analytics: pitch and torque deviation, component-temperature anomaly and peer comparison, and consolidated subsystem status.
  • Long-term degradation and Performance Loss Rate (PLR) trending.
  • Structured events and cases that feed the O&M workflow.

All of it runs unattended, per site, on native turbine data - no manual modelling step per asset.

What makes Performance+ unique

Highlight Why it matters
Self-learned 5PL power curve - a transparent 5-parameter-logistic curve fitted per turbine on density-corrected wind speed (IEC 61400-12-1), not a black-box regressor. Explainable, per-turbine reference that asset owners and OEMs can audit. See Wind Flagging Algorithm Reference.
Expected power with four methods + blend - contractual curve -> self-learned reference curve -> reference-turbine peers -> site mean, in strict precedence. Every turbine gets a defensible expected value even when its own signals are poor.
Capacity-normalised peer ("potential") power using only good-quality contributors, so mixed turbine models on one site can be pooled without bias. Robust "what the turbine should have made" that survives sensor faults and downtime.
Static and dynamic yaw misalignment - detects persistent yaw bias and grades how well each turbine dynamically follows the wind. Separates a fixed calibration offset from poor tracking dynamics - different fixes.
Comprehensive status detection - derate, curtailment (with control-changepoint and power-limit/setpoint context), abnormal operation, icing, thermal derating and wake loss, all per timestamp. Each lost interval is explained by a concrete, physical operating condition.
Icing detection on IEA Wind Task 19 P10/P90 reference curves built from the site's own SCADA. Quantifies icing loss without dedicated icing hardware.
Temperature-anomaly self-models trained only on genuinely-generating, clean samples, plus cross-turbine peer comparison of component temperatures. Early component-health warning that isn't fooled by idle periods or bad sensors.
Anemometer anomaly detection & daily sensor selection - rates wind and temperature sensors and picks the best signal to drive the analytics. Bad or drifting sensors no longer silently corrupt performance numbers. See Anemometer Anomaly Detection and Temperature Sensor Evaluation.
Long-term wind-speed correction & weather normalisation - MCP-style correction and Weibull observed-vs-long-term comparison feed a production waterfall. Separates a weak-wind year from genuine underperformance.
Foundational operational-state classification (idling / running / stopping / pitch- / torque-control) shared by every downstream flag. One consistent definition of "what the turbine was doing" across all analytics.
Standardised flagging architecture - every condition flag follows the same pipeline-safe contract. Consistent, testable behaviour across the whole detection layer. See Introduction to Wind Flagging Algorithms.

How it works - the pipeline at a glance

Performance+ executes as an ordered pipeline of steps, grouped into phases. Each phase consumes the outputs of the previous one, so data quality and modelling propagate cleanly downstream.

Phase Name What happens
0 Setup & validation Package-version logging, settings validation, point-similarity checks.
1 Input data Readiness check; ingests turbine SCADA, allocations, TMY, budget, satellite, temperature-tag and anomaly data.
2 Environment & market Contractual power-curve model, wake/sector geometry, budget downscaling, environmental data.
3 Data quality Per-object DQ for site, turbines, MET masts and meters; meter energy and communication status.
4 State & sensors Operational-state classification, sensor checks, sensor selection, weather imputation, air-density correction, relative wind direction.
5 Consolidation Consolidated turbine and site datasets (measured + modelled).
6 Power curves & expected power 5PL power-curve estimation, icing reference curves, expected power.
7 Turbine health Pitch & torque deviation, temperature self-models & peer comparison, subsystem status, health consolidation.
8 Status detection Derate, curtailment, abnormal operation, icing, thermal derating, wake loss, static & dynamic yaw, status consolidation.
9 Loss breakdown Per-turbine, site-aggregated, total and revenue loss breakdowns.
10 Results consolidation Consolidated per-turbine / per-site result set.
11 KPIs Turbine & site KPIs, degradation KPI and PLR (long- and short-term).
12 Root cause & events Underperformance root cause, event detection, cases workflow.
13 Long-term & waterfall Long-term wind-speed correction, Weibull observed vs long-term, weather adjustment, production waterfall.

Deep-dive technical notes

Performance+ results surface directly in the Bazefield interface - as KPIs, events and cases - giving asset managers and O&M teams one consistent, auditable view of plant performance.