专栏名称: AEii国际应用能源
发布应用能源领域资讯,介绍国际应用能源创新研究院工作,推广应用能源优秀项目,增进应用能源领域合作
目录
相关文章推荐
北京本地宝  ·  微信又有新功能上线! ·  3 天前  
51好读  ›  专栏  ›  AEii国际应用能源

【Applied Energy最新原创论文】基于联邦对比学习的中期天然气需求预测

AEii国际应用能源  · 公众号  ·  · 2023-07-06 20:41

正文

原文信息:

Federated deep contrastive learning for mid-term natural gas demand forecasting

原文链接:

https://www.sciencedirect.com/science/article/pii/S030626192300867X

Highlights

Propose a new paradigm to address data scarcity and diverse demand patterns problems.

Develop a new method to generate fine-grained representations from gas demand data.

Provide new findings for mid-term gas demand forecasting of different ranges.

摘要

准确的中期天然气需求预测对于天然气公司和政策制定者实现可靠的天然气供应计划、供应合同管理和高效运营起着至关重要的作用。然而,中期天然气需求预测面临着由于月度数据的收集频率较低所带来的数据匮乏问题,以及异构负荷模式所带来的挑战。本文通过结合联邦学习、深度对比学习和聚类方法,提出了一种联邦对比预训练—本地聚类微调(FedCon-LCF)的新范式。所提出的方法可以利用多家天然气公司的数据,以保护隐私的方式克服数据匮乏问题,并通过考虑异构负荷模式的本地聚类回归实现高性能预测。本文提出的面向预测的对比学习模型(FOCL)融合了改进的分层对比损失以及多尺度回归损失,可以有效地从时间序列提取有效信息并生成细粒度表征,为精确预测提供基础。本文所提方法在从中国11个城市的11家天然气公司收集的数据集上进行了评估,该数据集包括10种类别的17648个客户。提出的方法优于基准LSTM模型,在3个月前、6个月前、9个月前和12个月前的天然气需求预测中,MSE平均提高25.30%,MAE平均提高16.52%。

Abstr act

Accurate mid-term gas demand forecasting plays a crucial role for gas companies and policymakers to achieve reliable gas supply plans, supply contracts management, and efficient operation to meet the increasing gas demand. However, mid-term gas demand forecasting faces the problems of data paucity caused by the low frequency of collecting monthly data and heterogeneous consumption patterns of various usage categories. This paper proposes a novel Federated Contrastive pretraining - Local Clustered Finetuning paradigm (FedCon-LCF) by integrating federated learning, deep contrastive learning, and clustering approaches. The proposed method can utilize data from multiple gas companies to overcome data paucity issues in a privacy-preserving way, and high-performance forecasting can be achieved by local clustered regression considering the heterogeneous patterns. An improved hierarchical contrastive loss and multi-scale regression loss are integrated to develop the Forecasting-Oriented Contrastive Learning model (FOCL), which can effectively extract information and generate fine-grained representations of time series for accurate forecasting. The proposed method is evaluated on a dataset collected from 11 gas companies in 11 different Chinese cities with a total of 17648 clients over 10 usage categories. The proposed method outperforms the benchmark LSTM model with an average improvement of 25.30% in MSE and 16.52% in MAE for 3-month-ahead, 6-month-ahead, 9-month-ahead, and 12-month-ahead gas demand forecasting.

Keywords

Gas demand forecasting

Federated learning

Contrastive learning

Privacy-preserving

Heterogeneous consumption patterns

Graphics


Fig. 2. The structure of the proposed FOCL model, including data augmentation method, encoder structure, and loss function.

Fig. 3. The proposed federated contrastive pretraining - local clustered finetuning paradigm.

Fig. 4. The performance of the FedCon-LCF in comparison to the local benchmark on 413 clients. Different colors represent 3-month-ahead, 6-month-ahead, 9-month-ahead, and 12-month-ahead forecasting respectively. The x-axis shows the performance of the local LSTM, and the y-axis represents the performance of the FedCon-LCF model. The dots indicate that FedCon-LCF performs better than local LSTM in the area below the diagonal dashed line. (a) The model performance comparison in terms of MSE. (b) The model performance comparison in terms of MAE.

团队介绍







请到「今天看啥」查看全文