Switzerland's Energy Strategy 2050 promotes the use of renewable energy resources. Hydropower and solar energy peak in the summertime, while the demand for energy increases in the wintertime, which creates a seasonal mismatch between energy production and demand. Wind energy production peaks in the wintertime, thus making it a possible solution for satisfying the energy demand in Switzerland. However, the complex terrain topography, which comprises most of Switzerland's area, adds to the challenge, especially during the wind energy assessment. This study aims to enhance the efficiency and accuracy of the wind energy assessment in mountain regions through the combination of models and measurements.
Two measurement campaigns were conducted at the high altitude sites Lukmanier and Les Diablerets (hereafter Diablerets), Switzerland, using a near-surface ultrasonic anemometer and Light Detection and Ranging (LiDAR). Both campaigns were conducted for a short period of two and three months for the Lukmanier and Diablerets sites respectively. We improve the postprocessing of LiDAR measurement data to adapt to the complex terrain measurement challenges. After the post-processing steps, an Artificial Neural Network (ANN) is used to combine the wind measurement from LiDAR with the surrounding weather station data and obtain a long-term prediction of wind speed at a typical turbine hub height. A validation of the ANN shows that it successfully predicts the multi-year wind speed distribution.
To better understand the spatial variability of wind speed in complex terrain, we run a combination of the 'Consortium for Small-scale Modeling' and 'Weather Research and Forecasting' (COSMO-WRF) models initialized and driven by the COSMO-1E model. The Lukmanier and Diablerets simulation domains are centered at the LiDAR measurement locations and use a 300 m horizontal grid resolution. The impact of two mountain flow features, namely Foehn and mountain waves, on the wind energy potential in the Alps region is investigated. The results show a high probability of occurrence for both Foehn and mountain waves, thus emphasizing the need to consider complex terrain meteorological phenomena in a thorough and accurate wind energy assessment.
To estimate the wind energy potential over large areas and long periods, computationally efficient methods are needed. We compare the deep learning-based downscaling approach Wind-Topo with a variety of data (turbine measurements and LiDAR) and models (COSMO-1E, COSMO-WRF, and Wind Atlas Switzerland). Wind-Topo is shown to reduce biases from COSMO-1E, especially in high wind speed situations where COSMO-1E tends to overestimate compared to the measurement data. Moreover, Wind-Topo is shown to predict better the LiDAR-based wind speed measurement at the Lukmanier site in the condition of a relatively constant wind speed. The channeling effect is better captured in Wind-Topo due to the high topography resolution. Wind-Topo shows the ability to represent the high-resolution topography characteristics and increases the wind speed from COSMO-1E in areas such as side ridges and ridge summits, for examples.