摘要
针对抽水蓄能机组运行过程中数字化、智能化水平不足,以及现有仿真软件计算耗时较长、无法满足数字孪生系统实时性需求等问题,提出了一种基于数字孪生技术的解决方案,旨在提升对水泵水轮机的监控预测能力。采用本征正交分解降阶方法建立降阶模型,并通过FMI协议将其集成至系统中,结合Open3D点云实现了可视化交互,整个过程仅耗时728.6 ms;同时,基于长短期记忆网络算法开发了压力脉动预测模型,用于运行状况的提前预测,且预测的准确性大于96%;此外,通过Unity3D平台开发完整的数字孪生系统,并验证了其有效性和可行性。研究可为抽水蓄能机组的数字化、智能化提供一种全新的解决方案,对水力发电行业的智能化升级和现代化转型具有重要意义。
To address the insufficient digitalization and intelligence of pumped storage units during operation,as well as the long computation time of existing simulation software that fails to meet real-time requirements for digital twin systems,this study proposes a digital twin-based solution to enhance monitoring and prediction capabilities for pump-turbines.A reduced-order model was established using proper orthogonal decomposition(POD)and integrated into the system via the FMI protocol,achieving visual interaction through Open3D point cloud with the entire process taking only 728.6 milliseconds.Simultaneously,a pressure pulsation prediction model based on long short-term memory(LSTM)networks was developed for operational condition forecasting,with prediction accuracy exceeding 96%.Additionally,a complete digital twin system was developed using the Unity3D platform and validated for effectiveness and feasibility.This research provides a novel solution for the digitalization and intelligence of pumped storage units,holding significant implications for the intelligent upgrade and modern transformation of the hydropower industry.
作者
李琪飞
李润涛
辛路
陈祥玉
LI Qifei;LI Runtao;XIN Lu;CHEN Xiangyu(College of Energy and Power Engineering,Lanzhou University of Technology,Lanzhou 730050,China;Key Laboratory of Fluid Machinery and Systems,Gansu Province,Lanzhou 730050,China)
出处
《流体机械》
北大核心
2025年第5期102-109,共8页
Fluid Machinery
基金
国家自然科学基金项目(52066011)。
关键词
水泵水轮机
数字孪生
降阶模型
长短期记忆网络
监控预测
pump-turbine
digital twin
reduced-order model
long short-term memory network
monitoring and prediction
