Improvement of short-term numerical wind predictions

With the sustained growth of wind energy installed capacity for electricity generation, electricity system operators have increasing challenges balancing the electricity grid, notably in regards to minimizing the cost of other energy sources dispatch. Due to the variability of wind, wind power generation forecasting is an important issue for the economic viability of wind energy, whether in regulated or open markets. Therefore, there is a pressing need for robust short-term (up to 48 hours) surface wind forecast models, and eventually wind power forecast models, in order to sustain the integration of wind energy in electricity portfolios of jurisdictions. \n \nComputed for the needs of the wind energy industry, three years of experimental meteorological forecasts in Eastern Canada are available from Environment Canada Numerical Weather Prediction (NWP) model configured on a limited-area (GEM-LAM 2.5 km) for wind power predictions. These data include forecasts for the region of North Cape (Prince Edward Island) where the Wind Energy Institute of Canada runs a test site for wind turbines. Although the model spatial resolution is already relatively high (2.5 km), preliminary statistical analysis and site inspection revealed that the model does not have sufficient grid spacing refinement to properly represent the meteorological phenomena on this complex coastal site. For this reason, a Geophysic Model Output Statistic (GMOS) module has been developed and applied to optimize the use of the short-term NWP. GMOS differs from other MOS that are widely used by meteorological centers in the following aspects: 1) it takes into accounts the surrounding geophysical parameters such as surface roughness, terrain height, etc. along with the wind direction; 2) GMOS can be directly applied for model output correction without any training although a training of the GMOS will further improve the results. \n \nThis statistical module was trained and tested over the North Cape site and it basically improves the predictions RMSE by 25 – 30 % for all time horizons and almost all meteorological conditions. Also, the topographic signature of the forecast error due to insufficient grid refinement is eliminated and the NWP combined with the GMOS now outperforms the persistence model after a 2 h horizon, instead of 4 h without the GMOS. Ultimately, in order to generalize the results, this methodology has been validated by an independent test case performed on a site located in Bouctouche (New Brunswick). Similar improvements on the GEM-LAM 2.5km forecasts were observed thus, showing the general \napplicability of the GMOS. \n \nAlthough the current study presents an optimization of the use of short-term NWP for wind power forecasts using a statistical module, it also contributes to the development of a methodology and an analysis tool to assess and understand the NWP uncertainties on the amplitude and the phase of the surface wind forecast errors for different meteorological situations. A better knowledge of the wind speed and wind power forecast uncertainties, along with more accurate short-term wind forecast models, will increase the economic value of wind energy on the market.

Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.