WindMind provides state-of-the-art accuracy wind power production forecasts for asset owners, power traders and system operators. By combining weather data (such as wind speed, pressure, temperature, etc.) from numerical weather prediction (NWP) models, satellites and ground measurements with advanced machine learning methods that learn both local and global weather patterns. Processing of spatio-temporal information allows WindMind to produce forecast for individual wind turbines, utility scale wind farms and distributed region or country aggregation level generation with state-of-the-art accuracy. Compared to traditional forecasting methods, WindMind achieves a significant reduction in forecast errors by up to 20%, which lowers the imbalance costs from trading operations and helps improve system stability. The specific wind park characteristics are modeled implicitly by the machine learning method. Therefore, WindMind does not rely on static data, such as wind turbine manufacturer, wind farm layout, topology and vegetation, but learns the effect of these parameters from observed data. This results in higher accuracy than compared to physical models where uncertainty regarding static variables is translated to the forecast.
WindMind continuously incorporates the latest weather and power information to provide frequent forecast updates. In addition, performance on the first few hours ahead can be further reduced by adding live production data, although this is not a requirement. Through probabilistic power forecasts, the forecast uncertainty assessed. This gives higher degree of preparation for power market operations and allows for optimized intraday balancing of wind power. WindMind is provided as a cloud service and thus requires no installation, it can be easily integrated with your current system setup through API or SFTP-server.