Presentations

Abstract

Introduction As installed wind capacity grows, the industry increasingly faces the challenge of quantifying wake losses caused by neighboring wind farms. After the commissioning of a nearby wind farm, asset owners and operators often experience reduced power production, making it critical to redefine expected energy outputs. Assessing wakes losses, including contribution from nearby wind farms, is essential for characterizing wind farm performance, improving financial planning, and addressing potential compensation claims. Currently, two approaches stand out to evaluate external wake losses during operations. The first involves using energy yield assessment tools: using wake models, the expected increase in external wake losses can be estimated. However, this approach is limited by the potential mismatch between the calculated and actual wake effects during the period of interest. The second approach is based on the comparison between operational energy yield assessment results before and after the start of operation of the nearby farm, using models that predict power production given an independent wind reference. This requires several years of data to yield reliable results and can lead to uncertain conclusions if the time period includes other events leading to performance changes.  Method To overcome these limitations, we propose a novel method to determine operational wake losses more systematically, including contributions from neighboring wind farms, on a 10-minute timescale. The method consists of three key steps:  1. Estimating total wake loss timeseries using SCADA data: This step adapts a common approach [1] that estimates 10-minute wake power deficit timeseries by identifying turbines unaffected by turbine interactions to serve as the consensus freestream power reference. To extend this approach to scenarios with non-homogeneous inflow and mixed layouts, a wind flow model is used.  2. Calibrating an engineering wake model: The parameters of an engineering wake model are adjusted as in [2] but on a site specific basic to align the predicted wake losses with those observed from SCADA data. The wake model is also configured to account for neighboring wind farms.  3. Decomposing wake loss contributions from neighboring wind farms: The total wake loss timeseries is decomposed for each timestep to isolate the contributions of neighboring wind farms. This is achieved through side-by-side simulations using the wake model, enabling the relative contributions to be determined. Results can then be aggregated as desired, such as on a monthly basis.  Real world case studies The method is applied to three onshore sites, where operational data spanning several years was analyzed during periods when neighboring wind farms were commissioned. For validation, we evaluate the consistency of operational energy yield assessment results over the time period, both with and without accounting for wake losses from neighboring farms. Improvements in the regression coefficient are assessed to demonstrate the method's effectiveness.  [1]  Jacquet, C. Estimating wake losses in operating wind farms: Leveraging scada data and machine learning techniques. Wind Europe Annual Event, 2024.  [2]  Nygaard, NG et al. Large-scale benchmarking of wake models for offshore wind farms. In Journal of Physics: Conference Series, volume 2265, page 022008. IOP Publishing, 2022.