摘要
针对电力负荷预测中存在的随机性、不确定性的问题,结合深度学习算法具有很强的自适应感知能力等特点,采用目前较为主流的深度学习方法,如长短时记忆(LSTM)网络、门循环单元(GRU)神经网络和栈式自编码器(SAE),分别研究其应用于电力负荷预测时的效果。研究发现,将历史负荷数据作为三种深度学习预测模型的输入时,三种预测模型的负荷预测精度指标评估结果各有不同。因此,为了全面评估三种预测模型的预测效果,提出将不同时间段内的相同历史负荷数据作为预测模型输入对比各模型的负荷预测精度,从中找出最佳的预测模型。仿真结果验证了三种预测模型在电力负荷预测应用中的可行性,且发现在单输入因素时LSTM模型的预测精度相对较高。因此,在后续研究中,可以考虑以LSTM预测模型作为基础预测模型,结合更多的负荷影响因素进行改进,以提高负荷预测精度。
Aiming at the problems of randomness and uncertainty in power load forecasting,combining with deep learning algorithm,which has strong adaptive sensing ability,and by adopting the current mainstream deep learning methods,such as long-short-time memory (LSTM),gated recurrent unit (GRU) neural network and stacked automatic encoders (SAE),the prediction effects of power load forecasting are studied respectively.It is found in the study that when historical load data are used as input to three deep learning prediction models,the load forecasting accuracy indicators of the three prediction models have different evaluation results.In order to comprehensively evaluate the prediction effects of the three prediction models,the same historical load data in different time periods are used as the predictive model input,and the load prediction accuracy of each model are compared to find the optimal model.The simulation results verify the feasibility of three prediction models in power load forecasting applications,and find that the prediction accuracy of LSTM model is relatively high when single input factors are used.Therefore,in the follow-up study,the LSTM prediction model can be considered as the basic prediction model,combined with more load influencing factors to improve the load prediction accuracy.
作者
张建寰
吉莹
陈立东
ZHANG Jianhuan;JI Ying;CHEN Lidong(School of Aerospace Engineering,Xiamen University,Xiamen 361000,China)
出处
《自动化仪表》
CAS
2019年第8期8-12,17,共6页
Process Automation Instrumentation
基金
国家电网公司科技基金资助项目(JL71-16-006)
关键词
深度学习
长短时记忆
门循环单元
循环神经网络
栈式自编码器
负荷预测
预测精度
Deep learning
Long-short-term memory(LSTM)
Gated recurrent unit(GRU)
Recurrent neural networks (RNN)
Stacked auto-encoder(SAE)
Load forecasting
Prediction accuracy