Wind Energy Plugins for Weather Prediction Models

Numerous programmes and initiatives have recently been created to make high-fidelity weather and climate prediction data more easily accessible to the public. In an increasingly digital world, these predictions can be directly used to create products and services that have an enormous impact on communities, businesses, and people’s lives in general. One of the most prominent efforts in this area is Destination Earth, the European Union’s flagship initiative to develop Digital Twin models of the planet Earth. The European Centre for Medium-Range Weather Forecasts (ECMWF) contributes to the Destination Earth initiative by co-developing the software infrastructure that allows data to flow efficiently from the weather and climate models to the downstream applications.As part of this effort, ECMWF is developing a software called Plume that allows a weather model to extend its data processing functionalities through plugins. Plume plugins can read model data “on-the-fly“ (i.e. in memory) at each time step of the simulation. This is a key advantage as not all model data is available once the simulation has ended (data is in fact saved to disk at reduced frequency to avoid prohibitively expensive I/O operations and unmanageable data volumes). Plugins are also an effective way to deal with the complexity of large Weather Prediction models by allowing developers to implement data processing capabilities as well-defined, modular and more easily approachable software components. Therefore, plugins also offer great opportunities for collaborative development across institutions, third parties and scientific communities. In this context of collaborative development, ECMWF is also contributing to the EU Horizon project DTWO, that develops a Digital Twin for Wind Energy applications, making use of several data sources including Destination Earth.This work presents the development of Plume plugins for wind energy in the DTWO project, and more specifically, a wind farm modelling plugin and a weather extreme events detection plugin. The wind farm plugin reads wind fields from memory at every time-step and implements an algorithm to model wind turbine wakes and their interactions, in a pre-defined geographic area. By operating directly on in-memory data, the plugin provides access to high-frequency wind fields and allows a more granular prediction of wind turbine wakes and energy yield. The extreme-event detection plugin scans selected model fields and applies user-defined extreme-event detection algorithms. When events of interest are found, the plugin can also notify a server with information relative to the events. This mechanism can be used to trigger automatic data processing workflows in response to selected notifications. Finally, the functionalities developed in these plugins will be available to weather prediction models, creating synergies across projects and maximising the impact of this development.