摘要
针对区域电力负荷的时间序列数据随机性强、预测精度低及单一模型的数据特征提取能力差等问题,提出了一种支持向量机(SVM)、STL时序分解法、长短期记忆神经网络(LSTM)组合的电力负荷预测模型。该模型利用SVM对时间序列的电力负荷数据进行初始预测,并通过STL时序分解法对残差序列进行时序分解,从而提高残差序列的稳定性,减小其随机性,最后用LSTM对SVM的预测误差进行修正。试验结果证明,该方法利用误差修正可有效处理随机性强的数据,有利于预测结果的稳定性,提高预测精度。
Aiming at the problems of strong randomness of time series data of regional power load,low prediction ac-curacy and poor data feature extraction ability of a single model,a combined power load forecasting model based on sup-port vector machine(SVM),STL time series decomposition method,and long short-term memory neural network(LSTM)was proposed.This model uses SVM to initially predict the power load data of a time series,and uses STL time series decomposition to decompose the residual sequence,thereby improving the stability of the residual sequence and re-ducing its randomness.Finally,the LSTM is used to correct the prediction error of the SVM.The experimental results show that this method can effectively process highly random data using error correction,which is conducive to the stabili-ty of prediction results and improving prediction accuracy.
作者
王晨
李又轩
吴其琦
邬蓉蓉
WANG Chen;LI You-xuan;WU Qi-qi;WU Rong-rong(School of Automation,Guangxi University of Science and Technology,Liuzhou 545006,China;Guangxi Power Grid Limited Liability Company,Electric Power Research Institute,Nanning 530023,China)
出处
《水电能源科学》
北大核心
2024年第4期215-218,共4页
Water Resources and Power
基金
广西自然科学基金项目(2018GXNSFAA050029)。
关键词
组合模型
支持向量机
STL时序分解
长短期记忆网络
短期预测
误差修正
combined model
support vector machine
STL time series decomposition method
long short-term memory network
short-term prediction
error correction
