Data-driven stochastic diffusion process for wind speed modelling and forecasting

A novel hybrid model, which integrates a stochastic diffusion process and artificial neural networks, is proposed for accurate modelling and forecasting of wind speed. A stochastic diffusion process model is deployed to capture the uncertainty of wind speed, while the drift and the diffusion of the stochastic diffusion process are approximated by artificial neural networks (ANNs). The proposed method takes advantages of both the stochastic diffusion process to simulate the evolution of wind speed and the universal approximation of ANNs to approximate the distribution of wind speed. The proposed method can easily be used to generate wind speed at any length. The effectiveness of the proposed method is validated through numerical studies on wind speed data collected from National Renewable Energy Laboratory, USA, by benchmarking against the persistence model and the improved Markov chain model in the literature. The numerical results illustrate that the proposed method achieved better modeling and forecasting in terms of mean absolute error and root mean square error than the benchmarks.