Super-short-Time Wind Power Forecasting Based on EEMD-IGSA-LSSVM

Abstract

In order to improve the prediction accuracy of the output power of the wind farm under the premise of ensuring safe operation, a combination of wind power forecasting model based on Ensemble Empirical Mode of Decomposition (EEMD), Improved Gravitational Search Algorithm (IGSA) and Least Squares Support Vector Machine (LSSVM) was established. Firstly, the wind power time series was decomposed into a series of subsequences with significant differences in complexity by using EEMD algorithm. Secondly, the decomposed subsequence was reconstructed by the phase space reconstruction (PSR), and then, an IGSA-LSSVM prediction model of each sub-sequence reconstructed was established respectively. In order to analyze the differences of LSSVM which sets up different kernel functions, eight kinds of kernel function LSSVM prediction models were established, and the IGSA algorithm was adopted to solve those models. Finally, taking a wind farm in Inner Mongolia of China as an example, the simulation and calculation results illustrate that LSSVM prediction model based on the exponential radial basis kernel function and penalty factor obtained by using the IGSA algorithm has higher prediction accuracy. Compared with five conventional combined models such as EMD-WNN and EMD-PSO-LSSVM, the combined model EEMD-IGSA-LSSVM of exponential radial basis kernel function mentioned above can forecast wind power in an effective and accurate way.

Keywords: ensemble empirical mode decomposition (EEMD),  wind power prediction,  least squares support vector machine (LSSVM),  improved gravitational search algorithm(IGSA),  exponential radial basis function(ERBF)