Integrative Density Forecast and Uncertainty Quantification of Wind Power Generation

The volatile nature of wind power generation creates challenges in achieving secure power grid operations. It is, therefore, necessary to accurately predict wind power and its uncertainty quantification. Wind power forecasting usually depends on wind speed prediction and the wind-to-power conversion process. However, most current wind power prediction models only consider portions of the uncertainty. This paper develops an integrative framework for predicting wind power density, considering uncertainties arising from both wind speed prediction and the wind-to-power conversion process. Specifically, we model wind speed using the inhomogeneous Geometric Brownian Motion and convert the wind speed prediction density into the wind power density in a closed-form. The resulting wind power density allows quantifying prediction uncertainties through prediction intervals. To forecast the power output, we minimize the expected prediction cost with (unequal) penalties on the overestimation and underestimation. We show the predictive power of the proposed approach using data from multiple operating wind farms located at different sites.

I. Introduction

Unlike traditional fossil-based energy sources, wind power generation is severely affected by stochastic weather conditions [1], posing significant challenges in achieving secure power grid operations [2]. Thus, accurate forecasting of wind power generation and its uncertainty quantification becomes a critical component in several decision-making processes including unit commitment, economic dispatch, and reserve determination [3]. Wind power generation forecasts have been widely investigated in the literature (e.g., [4], [5]). Interestingly, many studies focus on generating point forecasts of wind power. However, due to the highly volatile and intermittent nature of wind power, probabilistic density forecasts become more crucial for decision-making, e.g., energy storage system sizing, in power system operations under large uncertainties [4], [6].