Actual experience on the short-term wind power forecasting at Penghu — From an island perspective

Wind power is expected to contribute significantly to the renewable energy targets owing to advancements in wind technologies, falling capital costs, abundance of free resource and commercial viability. However, it also brings into a lot of new challenges for the power network planning and operation. The predictability of wind power in managing load and generation balance is very important. If the fluctuations of wind were perfectly predictable, the additional cost of operating the system with a large penetration of wind power would be reduced. Therefore, accurate and reliable forecasting systems for wind power are widely recognized as a major contribution for increasing wind penetration. The target of this paper is to investigate the state-of-the-art wind power forecast technologies all over the world and predict wind power generated by the Penghu Jhong-tun wind farm. The short termand very short term forecasts, including ten-minute-ahead, one-hour-ahead, and one-day-ahead predictions, have been taken into account respectively. This prediction module will be developed with statistical models and Artificial Intelligence (AI) technologies respectively to capture the relations between input variables, i.e., online measurement data and numerical weather prediction (NWP) based on meteorological information, and the output valuable, such as future total wind farm power. Furthermore, the significance of NWP values on wind power forecasting has been highlighted in this paper by using a real example at the Penghu Jhong-tun wind farm.