Presentations

Abstract

QUANTIFYING WIND SPEED MEASUREMENT AND PRODUCTION LOSS CALCULATION RELIABILITY DURING TURBINE DOWNTIME: A 2200 TURBINE-WIDE STUDY Reliable wind speed estimation is the cornerstone of accurately assessing energy production losses during turbine downtime. In utility-scale wind farm operations, "loss attribution" is a high-stakes financial activity. Precise calculations are essential for settling insurance claims, verifying availability guarantees in Service Level Agreements (SLAs), and identifying underperforming assets to prioritize maintenance. However, standard industry practices for calculating these losses often rely on assumptions that fail to reflect real-world operational complexities. When a turbine is offline, its own nacelle-mounted sensors become unreliable due to altered aerodynamic profiles and the lack of active yaw alignment. Beyond this, other common industry methods introduce significant errors: * Neighboring Production Proxies: Using power output from adjacent turbines often overlooks micro-siting differences, local wake effects, and varying performance levels between individual units. * Theoretical OEM Power Curves: Relying on manufacturer-provided curves is frequently inaccurate because turbines in the field rarely mirror "ideal" wind-tunnel conditions due to blade degradation, site-specific turbulence, and varying control settings. This study addresses these gaps by conducting a massive statistical analysis of wind speed and production dynamics. The dataset is one of the most comprehensive ever analyzed for this purpose, comprising over 2,200 wind turbines across 150+ wind farms. The study covers a total installed capacity of 6.5 GW, encompassing 10 different Original Equipment Manufacturers (OEMs) and 50 unique turbine models. By utilizing 10-minute granular SCADA data with an average of three years of history per turbine, the analysis represents 6,600 turbine-years of data. This scale allows us to account for a wide array of confounding features, including air density variations, specific wind regimes (high-turbulence vs. laminar), turbine-specific power production performance, and make/model characteristics. The research focuses on quantifying the shift in spatial correlation dynamics during non-operational periods. We systematically examine the error margins introduced when current industry "shortcuts"—such as using raw nacelle wind speed or theoretical curves—are applied to different wind regimes and turbine types. Rather than proposing a single "correct" proxy, this study highlights the conditions under which traditional methods fail and identifies the specific features (e.g., air density and local wake dynamics) that most significantly impact the reliability of a downtime loss calculation. Our findings provide a evidence-based framework for asset managers and insurers to move away from "one-size-fits-all" calculations toward more robust, site-specific diagnostics. By demonstrating the non-linear relationship between downtime conditions and measurement reliability, this work contributes to the development of more transparent contractual standards and improved financial modeling in the global wind energy sector.