Frequency Regulation Potential Prediction of Wind Power Based on Historical Data

Abstract

With the rapid growth of wind power inertial response and primary frequency control technology, the potentialities of wind farm frequency regulation are required to be accurately quantified and predicted. To be able to reflect the multi-time scale characteristics of frequency response, this paper aims to analyzing the main factors affecting the performances of the frequency regulation and proposes a novel prediction method to improve the accuracy of potential predictions based on historical data. The wind speed time series is decomposed by improved complete ensemble empirical mode decomposition (ICEEMD) and singular spectrum analysis (SSA), and then the least squares support vector machine (LSSVM) is used to make multi-step predictions for each component. The predicted wind speed results are used to characterize the wind power frequency regulation potential under different frequency control methods. The research results show that the proposed ICEEMD-SSA-LSSVM method can improve the prediction accuracy and realize the precise prediction of the wind power frequency regulation potential.