Aggregated Probabilistic Wind Power Forecasting Based on Spatio-Temporal Correlation

In this paper, an aggregated probabilistic wind power forecasting method by considering spatio-temporal effects is proposed, which consists of three major components: Q-learning enhanced deterministic forecasting, marginal distribution fitting, and aggregated probabilistic wind power forecasting. A high-dimensional joint distribution is used to model the spatiotemporal correlation among the member wind farms through Copula, where the marginal distributions are built from historical aggregated wind power observations and forecasts from member wind farms. Then, a conditional distribution of the aggregated wind power is deduced through the Bayesian theory, which is used for aggregated probabilistic forecasts. Numerical results of a 3 wind farms case study show that the performance of the aggregated probabilistic wind power forecasting is enhanced by considering the spatio-temporal correlation among individual wind farms.