首页|基于LSTM-XGBoost和多模型算法的短期负荷预测

基于LSTM-XGBoost和多模型算法的短期负荷预测

扫码查看
针对负荷数据波动性强、特征存在冗余而导致使用单一模型预测短期负荷时精度较低的问题,提出一种融合梯度提升树(GBDT)、自适应噪声完备集合经验模态分解(CEEMDAN)、长短期记忆(LSTM)和极端梯度提升(XGBoost)的短期负荷预测组合方法.首先利用GBDT对负荷数据集进行特征选择,筛选出重要特征;然后使用CEEMDAN将负荷序列分解后合并为低频分量和高频分量;再将低频分量输入到LSTM中进行预测,将高频分量输入到XGBoost中进行预测;最后,短期负荷的最终预测结果由两个模型的预测结果进行叠加而成.与单一预测模型相比,所提方法在短期负荷方面具有更高的准确性.
Short-term load predicting based on LSTM-XGBoost and multi-model algorithm
To address the problem of low accuracy in predicting short-term load using a single model due to the high volatility of load data and redundancy of features,a combined method for short-term load prediction is proposed that combines GBDT,CEEMDAN,LSTM,and XGBoost.Firstly,GBDT is used to feature select the load dataset to filter out important features.Next,the load sequence is decomposed and merged into low-frequency and high-frequency components using CEEMDAN.Then,the low-frequency components are input into LSTM and the high-frequency components are input into XGBoost for prediction.Finally,the final prediction results of short-term load are made by superposing the prediction results of the two models.Compared with the single prediction model,the proposed prediction method has higher accuracy in the short-term load.

long short-term memory(LSTM)extreme gradient boost(XGBoost)short-term load predictioncomplete ensemble empirical mode decomposition with adaptive noise(CEEMDAN)gradient boosting decision tree(GBDT)

邵必林、庄雪莉、曾卉玢

展开 >

西安建筑科技大学管理学院,陕西 西安 710055

长短期记忆 极端梯度提升 短期负荷预测 自适应噪声完备集合经验模态分解 梯度提升树

国家自然科学基金面上项目

62072363

2023

计算机时代
浙江省计算技术研究所 浙江省计算机学会

计算机时代

影响因子:0.411
ISSN:1006-8228
年,卷(期):2023.(12)
  • 10