电力系统及其自动化学报2025，Vol.37Issue(9)：34-44,11.DOI:10.19635/j.cnki.csu-epsa.001604
改进贝叶斯优化与集成学习短期负荷预测模型
Short-term Load Forecasting Model Based on Improved Bayesian Optimization and Ensemble Learning
摘要
Abstract
A forecasting model based on improved Bayesian optimization and ensemble learning is proposed for short-term load forecasting in power systems.First,the ridge regression,least absolute shrinkage and selection operator(LASSO)regression,random forest(RF)and Huber regression are used as base learners to independently predict the target load value,with the load data in previous periods as independent variables.Second,an extreme gradient boosting(XGBoost)-based meta learner is constructed,and the predictions from base learners are used as independent variables to further improve the prediction accuracy of the target load value.Third,a Hyperband-based improved Bayesian optimi-zation algorithm and 5-fold cross-validation are employed for parameter optimization.Finally,shapley additive explana-tions(SHAP)values are used to analyze the importance of each base learner to the meta learner,as well as the impor-tance of each input feature to the ensemble learning model.Simulation results and a case study demonstrate that the pro-posed method outperforms single models in terms of prediction accuracy and stability.Compared with neural networks,it improves the prediction interpretability without sacrificing the prediction accuracy.关键词
电力系统/短期负荷预测/贝叶斯优化/集成学习/极致梯度提升/沙普利加和解释值Key words
power system/short-term load forecasting/Bayesian optimization/ensemble learning/extreme gradient boosting(XGBoost)/shapley additive explanations(SHAP)value分类
信息技术与安全科学引用本文复制引用
顼佳宇,王晓冰,李冰,王媛,雍明月,邵晨..改进贝叶斯优化与集成学习短期负荷预测模型[J].电力系统及其自动化学报,2025,37(9):34-44,11.基金项目
国网北京市电力公司科技项目(520234240002). （520234240002）
