| 引用本文: | 王继东,杜冲.基于Attention-BiLSTM神经网络和气象数据修正的短期负荷预测模型[J].电力自动化设备,2022,42(4): |
| WANG Jidong,DU Chong.Short-term load prediction model based on Attention-BiLSTM neural network and meteorological data correction[J].Electric Power Automation Equipment,2022,42(4): |
|
| 摘要: |
| 基于循环神经网络的负荷预测模型大多将历史负荷数据和影响负荷的其他因素如气象数据等共同作为预测模型的输入特征，但气象数据内部规律性不强，不适合作为循环神经网络的输入。针对该问题，提出一种基于Attention-BiLSTM神经网络和气象数据修正的短期负荷预测模型。采用最大信息系数分析影响负荷的主要因素；考虑到负荷序列较长且存在双向信息流，利用BiLSTM神经网络进行预测；引入注意力机制，通过注意力权重突出关键因素的影响，挖掘负荷数据的内部规律；利用核极限学习机结合气象数据进行误差预测和修正，完成负荷预测。以我国东部某地区真实数据作为实际算例，实验结果表明，所提模型与其他模型相比有更好的预测效果。 |
| 关键词: 短期负荷预测 最大信息系数 注意力机制 双向长短时记忆神经网络 核极限学习机 |
| DOI：10.16081/j.epae.202112017 |
| 分类号:TM715 |
| 基金项目:国家重点研发计划项目(2016YFB0901102) |
|
| Short-term load prediction model based on Attention-BiLSTM neural network and meteorological data correction |
|
WANG Jidong, DU Chong
|
|
Key Laboratory of Smart Grid of Ministry of Education, Tianjin University, Tianjin 300072, China
|
| Abstract: |
| Most of the load prediction models based on recurrent neural network take historical load data and other factors such as meteorological data together as input features of prediction model, but the internal regularity of meteorological data is poor, so it is not suitable for the input of recurrent neural network. Aiming at the problem, a short-term load prediction model based on Attention-BiLSTM(Attention Bi-directional Long Short-Term Memory) neural network and meteorological data correction is proposed. The maximum information coefficient is adopted to analyze the main factors that affecting the loads. Considering that the load sequence is long and there exists bi-directional information flow, BiLSTM neural network is used for prediction. The attention mechanism is introduced, the influence of key factors is highlighted by the weight of attention, and the internal regularity of load data is excavated. The kernel extreme learning machine is used for error prediction and correction combined with meteorological data, and the load prediction is completed. The real data of a certain area in eastern China is taken as practical example, and the experimental results show that the proposed model has better prediction effect than other models. |
| Key words: short-term load prediction maximum information coefficient Attention mechanism bi-directional long short-term memory neural network kernel extreme learning machine |
