An intelligent method for very-short range multi-step wind power forecasting

This paper proposes the development of very-short range multi-step wind power forecasting model based on functional network (FN), a modern intelligent paradigm. Although FNs are a well-developed form of neural networks, but the use of these models in renewable power forecasting is a new and emerging concept. The inherent architecture of FN offers problem-driven network topologies and optimal neural functions with various mathematical structures as opposed to classical neural networks. These advantages of functional networks produce a high-performance wind power forecasting model which is further validated in comparison with a benchmark model as well as a conventional neural network model for very-short range multi-step wind power forecasting. The results obtained through a real-world case study indicate notable improvement in forecast accuracy in terms of standard performance indices. Hence the proposed FN forecast model can become a useful tool for wind power system operators in multiple aspects of power system planning and dispatch.