基于多目标和贝叶斯优化的短期负荷区间预测

Short-term load interval forecasting based on multi-objective and bayesian optimization

新能源并网后的供电体系存在具有较高的间歇性和随机性,这将为电力生产和调度的平衡带来巨大挑战,而如何量化电力负荷的不确定性对电力系统安全经济地运行起着重要作用。为此,提出一种基于多目标和贝叶斯优化(multi-objective optimization and Bayesian optimization,MOBO)的深度学习区间预测模型,能在给定的置信水平下描述电力负荷的变化趋势。在预测模型的构建过程中,依据分位数回归理论计算出电力负荷在不同分位点处的预测区间,再通过有效性检验,筛选出合理的预测模型。同时,采用多目标优化和贝叶斯优化算法理论对深度学习模型的超参数进行调优。使用美国纽约州米尔伍德的电力负荷数据集对所提出的模型进行验证,实验结果表明,与其他模型相比,模型在不同置信水平下均有着更高的预测区间覆盖率和更窄的区间平均宽度,更能精确地描述未来电力负荷的波动范围。

The grid connected power supply system of new energy has high intermittency and uncertainty,which will bring great challenges to the balance of power production and dispatching.How to quantify the uncertainty of power load plays an important role in the safe and economical operation of power system.Therefore,this paper proposes a deep learning interval prediction model based on multi-objective optimization and Bayesian optimization(MOBO),which can describe the variation trend of power load at a given confidence level.In the process of building the prediction model,we calculated the prediction interval of power load at different points according to the quantile regression theory,and then screened the reasonable prediction model through the validity test.At the same time,multi-objective optimization and Bayesian optimization algorithm theory are used to tune the hyperparameters of the deep learning model.In this paper,the power load dataset of Millwood,New York,USA is used to verify the proposed model.The experimental results show that the proposed model has greater prediction interval coverage probability and smaller prediction interval normalized average width at different confidence levels than other models,besides more accurately describes the fluctuation range of future electricity load.