Publication

Abstract : The global priority in the production of clean and sustainable energy necessitates a reliable method of generation forecasts. However, the stochastic nature of renewable energy resources develops uncertainty in the decision-making processes of energy markets, which are not addressed in the traditional deterministic forecasts. This paper proposes a two-step process to implement a data-driven deterministic forecast-based probabilistic method for the potential quantification of uncertainty involved in wind power generation forecasts. In the first step, deterministic forecast approach integrates variational mode decomposition (VMD), discrete wavelet transforms (DWT) and Autoregressive integrated moving average (ARIMA) that creates a new combination of hybrid method for an effective short-term wind power forecast. The second step employs Gauss–Newton regression method to create an appropriate confidence interval (CI) and prediction intervals (PI) to assess the market risk allied with the uncertainty in wind power deterministic forecasts. CIs and PIs for the various confidence levels of 95%, 90% and 70% are constructed on the outcomes of the deterministic forecast using the Gauss–Newton regression method. Thereby the proposed two-step process approach presents a novel method for short-term wind power forecasts together with the uncertainty related to the forecasts. Testing with the historical wind speed data from two wind sites shows that VMD-DWT-ARIMA outperforms all other comparison models and Gauss–Newton regression method produces more rational CIs and PIs than the state-of-the-art methods.