%0 Journal Article
%T 基于CEEMDAN-GWO-KELM模型对我国电力需求预测<br>Electricity Demand Forecasting in China Based on CEEMDAN-GWO-KELM Model
%A 王斯
%A 张国浩
%A 杨澈洲
%J Computer Science and Application
%P 2073-2083
%@ 2161-881X
%D 2022
%I Hans Publishing
%R 10.12677/CSA.2022.129210
%X <span style=\"font-size:14px;\">中长期电力预测是电力系统规划主要研究之一，也是学界和业界广泛关注的热点之一。本文构建合理的电力需求影响因素指标体系，通过稀疏贝叶斯学习和相关性分析筛选出关键性指标。利用自适应噪声完备经验模态分解(Complete Ensemble Empirical Mode Decomposition with Adaptive Noise, CEEMDAN)将电力需求数据分解成多个信号分量(intrinsic mode function, IMF)，并将其作为待预测分量。利用灰狼优化算法(Grey Wolf Optimizer, GWO)对核极限学习机(Kernel Based Extreme Learning Machine, KELM)的参数进行优化，建立了CEEMDAN-GWO-KELM多重组合模型。通过1960~2020年的电力需求数据做模型对比的实证分析，验证了该模型的有效性，并对我国中长期电力需求进行预测。</span><br />
<span style=\"font-size:14px;\">Medium- and long-term power forecasting is one of the main studies in power system planning, and one of the hot spots widely concerned by academia and industry. In this paper, we construct a reasonable index system of power demand influencing factors, and screen out key indexes by sparse bayesian learning (SBL) and correlation analysis. Using complete ensemble empirical mode decomposition with adaptive noise (CEEMDAN), the electricity demand data are decomposed into multiple signal components (intrinsic mode function, IMF) and the IMF is used as the component to be predicted. The parameters of the kernel based extreme learning machine (KELM) are optimized by the grey wolf optimizer (GWO), and a CEEMDAN-GWO-KELM multiple combination model is developed. The empirical analysis of model comparison is done through the electricity demand data from 1960 to 2020 to verify the validity and accuracy of the model and to forecast the medium and long-term electricity demand in China.</span>
%K 中长期电力，电力需求，稀疏贝叶斯学习，模态分解，核极限学习机<br>Medium and Long-Term Electricity
%K Electricity Demand
%K Sparse Bayesian Learning
%K Modal Decomposition
%K Kernel Extreme Learning Machine
%U http://www.hanspub.org/journal/PaperInformation.aspx?PaperID=55541