摘要
随着建筑行业对能源需求的持续增长,传统的能耗预测模型在处理大规模数据集时面临性能瓶颈。文章探讨了如何利用高性能计算加速CNN-LSTM模型和注意力机制的集成,以提高建筑能耗预测的准确性和处理速度。通过并行计算和GPU加速,此研究显著提升了模型在特征提取和长期依赖分析上的效率。研究结果旨在为建筑能源管理提供更精确的数据支持,促进能效优化和环境影响的减少。
As the energy demand in the building industry keeps growing,traditional energy consumption prediction models encounter performance bottlenecks when handling large-scale datasets.This paper focuses on using high-performance computing to accelerate the integration of the CNN-LSTM model and attention mechanism,enhancing the accuracy and processing speed of building energy consumption prediction.Through parallel computing and GPU acceleration,this study significantly improves the model’s efficiency in feature extraction and long-term dependence analysis.The results aim to provide more accurate data support for building energy management and promote energy efficiency optimization and environmental impact reduction.
作者
程瀛
温锦成
陈明
张海滔
CHENG Ying;WEN Jincheng;CHEN Ming;ZHANG Haitao(Guangzhou Construction Co.,Ltd.,Guangzhou 510288,China)
基金
广州市建筑集团有限公司科技计划项目《专题八:广东省施工仿真和BIM工程技术研究中心建设》(项目编号:[2021]-KJ027)。
关键词
高性能计算
CNN-LSTM模型
注意力机制
建筑能耗
high performance computing
CNN-LSTM model
attention mechanism
building energy consumption
