 In island grids, grid operators face the challenge of curtailing wind energy production in order to maintain system stability and security of supply. Without storage or other forms of flexibility, such as demand side management, wind park owners must reject excessive amounts of wind energy throughout the year. This can be mitigated through the use of short-term leases of energy storage and or direct bilateral power purchase agreements with flexible demand entities. To facilitate these options, accurate wind energy forecasting is needed, both theoretically and practically. We have developed artificial neural network models for day-ahead forecasting of hourly granularity, tested them in a large-scale non-interconnected island system, and found that they provide a fair accuracy of day-ahead wind energy predictions. These results suggest that alternative actor schemes could emerge in similar systems, providing new opportunities for wind park owners. This article was authored by Konstantinos Moustres and Demetrius Zafferakis. We are article.tv, links in the description below.