AWESPA: a modular toolchain for assessing airborne wind energy systems and farms

Abstract

Airborne wind energy systems can access wind resources at higher altitudes while using less material than conventional wind turbines. Their performance assessment, however, requires several connected modeling steps, including wind resource analysis, power estimation, farm operation, and annual energy production calculation. These steps are often implemented in separate tools with different input and output formats, which makes results difficult to compare and analyses difficult to repeat. This thesis develops AWESPA, a modular Python toolchain for airborne wind energy system and farm performance assessment, together with awesIO, a standardized and validated input and output framework for airborne wind energy data.

The toolchain is built around exchangeable modules for each step in the performance assessment. The awesIO files provide common data definitions for the main information exchanged between modules and can be checked against predefined schemas to detect missing parameters, inconsistent file structures, and incorrect data definitions. The wind resource module converts ERA5 wind data into representative vertical wind profiles using principal component analysis and k-means clustering. A comparison with lidar measurements at the Bangor Erris test site shows that ERA5 captures the first-order wind climate, including the mean profile shape, Weibull scale parameter, and dominant wind direction. However, ERA5 underrepresents local wind-speed variability and produces a narrower range of profile shapes than the lidar data, especially for high-shear, low-shear, and low-level-jet-like profiles.

Three power estimation models are implemented and compared: the original Luchsinger model, the soft-kite Luchsinger model, and an inertia-free quasi-steady model. The Luchsinger models are useful for testing the modular structure of AWESPA, but give more optimistic power estimates than the inertia-free quasi-steady model. The difference is mainly caused by the treatment of gravitational loading and the reel-in phase. The quasi-steady model is therefore used for the case-study power estimates and is extended with variable reel-out elevation angles, a reel-out to reel-in transition, and constrained power-curve optimization.

At the farm level, a geometric maximum phase offset is derived for closely spaced wind-aligned systems with overlapping reel-in and reel-out envelopes. The results show that wind-aligned systems do not always need to operate synchronously. Limited asynchronous operation can reduce the peak-to-peak variation of the aggregated farm power output while maintaining geometric separation.

Two case studies are considered: a $100~\mathrm{kW}$ isolated system at Bangor Erris and a $1~\mathrm{MW}$ farm at Maasvlakte 2. The first case study compares system design choices and evaluates the selected system using site-specific wind profiles, giving an annual energy production of $423.3~\mathrm{MWh}$ with the clustered wind-profile representation. The second case study applies the same system in a 10-system farm and gives an annual energy production of $3690~\mathrm{MWh}$ with the clustered wind-profile representation. Together, the case studies show how AWESPA can be used to assess different sites, system configurations, and farm operating strategies in a consistent and reusable way.


https://github.com/awegroup/awesIO

https://github.com/awegroup/AWESPA