摘要
电力负荷预测是实现电力系统智能化的基础。准确的负荷预测可以保证电力系统安全稳定地运行。针对短期负荷波动大,随机性强的特点,提出一种基于极端梯度提升和时间卷积网络的短期电力负荷预测方法。首先,利用变分模态分解(variational mode decomposition,VMD)对负荷序列进行平稳化预处理,将原始负荷拆分成若干个模态分量负荷。同时,为了减少预测模型训练所需的时间,利用样本熵来评估各分量的复杂度,将复杂性相近的负荷分量归为一类用于训练同一模型。最后,结合极端梯度提升和时间卷积网络的优点,利用极端梯度提升模型来预测趋势负荷,时间卷积网络模型来预测随机扰动负荷,并在模型训练过程中,利用树状Parzen估计来调节模型的超参数,得到最优的预测模型。为验证本文所提方法的有效性,在EUNITE竞赛数据集上进行了仿真实验,分别预测未来24 h的短期负荷和每日峰值负荷。实验结果表明,相比于支持向量回归(support vector regression,SVR),长短时记忆(longshort-term memory,LSTM),门控循环单元(gated recurrent unit,GRU),经验模态分解(empirical mode decomposition,EMD)+LSTM等短期负荷预测模型,该方法能取得更好的预测效果,具有更高的预测精度。
Power load forecasting is the basis for achieving intelligent power system.Accurate load forecasting can ensure the safe and stable operation of the power system.Aiming at the short-term load characteristics of strong randomness and large fluctuations,we propose a short-term power load forecasting method based on extreme gradient boosting and temporal convolutional network.First,variational mode decomposition(VMD)is used to decompose the load sequence into multiple modal component loads for smoothing processing.In addition,in order to reduce the time consumed for model training,sample entropy is used to evaluate the complexity of each component,and the load components with similar complexity are classified into one category for training the same model.Finally,in combination with the advantages of extreme gradient boosting(XGBoost)and temporal convolutional network(TCN)models,XGBoost is used to predict trend load,and TCN is used to predict random disturbance load.In the process of model training,tree Parzen estimation(TPE)is used to adjust the hyperparameters of the model to get the best prediction model.In order to verify the effectiveness of the method proposed in this paper,simulation experiments were carried out on the EUNITE competition data set to predict the next 24 hours load and daily peak load,respectively.Experimental results show that,compared with short-term load forecasting models such as support vector regression(SVR),long short-term memory(LSTM),gated recurrent unit(GRU),and empirical mode decomposition(EMD)+LSTM,the proposed method can be adopted to achieve better forecasting results.
作者
唐贤伦
陈洪旭
熊德意
张艺琼
蒋维弛
邹密
TANG Xianlun;CHEN Hongxu;XIONG Deyi;ZHANG Yiqiong;JIANG Weichi;ZU Mi(Chongqing Key Laboratory of Complex Systems and Bionic Contro,Chongqing University of Posts and Telecommunications,Chongqing 400065,China;Chongqing Academy of Metrology and Quality Inspection,Chongqing 401123,China;State Grid Chongqing Yongchuan Power Supply Company,Chongqing 402160,China)
出处
《高电压技术》
EI
CAS
CSCD
北大核心
2022年第8期3059-3067,共9页
High Voltage Engineering
基金
国家自然科学基金(61673079)
重庆市自然科学基金(cstc2020jcyj-msxmX0693)。
关键词
负荷预测
变分模态分解
时间卷积网络
极端梯度提升
树状Parzen估计
load forecasting
variational mode decomposition
temporal convolutional network
extreme gradient boosting
tree Parzen estimation
