BIGDATA: IA: Collaborative Research: From Bytes to Watts - A Data Science Solution to Improve Wind Energy Reliability and Operation

BIGDATA：IA：协作研究：从字节到瓦特 - 提高风能可靠性和运行的数据科学解决方案

• 批准号: 1741173
• 负责人: Yu Ding
• 金额: $ 74.98万
• 依托单位: Texas A&M Engineering Experiment Station
• 依托单位国家: 美国
• 项目类别: Standard Grant
• 财政年份: 2017
• 资助国家: 美国
• 起止时间: 2017-10-01 至 2022-09-30
• 项目状态: 已结题
• 来源: https://www.nsf.gov/awardsearch/showAward?AWD_ID=1741173&HistoricalAwards=false
• 关键词: BIGDATA; IA; Collaborative; Research; Bytes

The collective efforts in aerospace, civil, electrical, and mechanical engineering areas have led to remarkable progresses in wind energy. Larger turbines are designed and installed, and wind farms are nowadays built at locations where wind is even more intermittent and maintenance equipment is less accessible. This adds new challenges to ensuring operational reliability. To cope with these challenges, along with the rapid advancement in microelectronics, modern wind farms are equipped with a large number and variety of sensors, including, at the turbine level, anemometers, tachometers, accelerometers, thermometers, strain sensors, and power meters, and at the farm level, anemometers, vanes, sonars, thermometers, humidity meters, pressure meters, among others. It is worth noting that all these data are currently analyzed/utilized only in their respective domains. The big data challenges in this project include how to best use spatio-temporal data for wind forecast, how to use data of different nature (wind, power, load etc.) and data of different sources (physical data versus computer simulation data) for power production assessment in a computationally efficient manner, and finally how to integrate these three sets of solutions into a reliable and efficient computational platform. The proposed research and education activities will make a paradigm shift in the wind industry by demonstrating how dramatically data science innovations can benefit the industry. The PIs will disseminate the research findings through classroom teaching, journal/conference publications, industry workshops, and data/software sharing. The summer internship opportunities and undergraduate research help train the next generation workforce to be better versed with data science methodologies.The critical barrier to cost effective wind power and its general adoption is partly rooted in wind stochasticity, severely complicating wind power production optimization and cost reduction. The long-term viability of wind energy hinges upon a good understanding of its production reliability, which is affected in turn by the predictability of wind and power productivity of wind turbines. Furthermore, the productivity of a wind turbine comprises two aspects: its ability of converting wind into power during its operation and the availability of wind turbines. Three inter-related research efforts will enhance wind energy reliability and productivity): (1) spatio-temporal analysis (for wind forecast) (2) conditional density estimation (for wind-to-power conversion assessment); and (3) importance sampling (for turbine reliability assessment and improvement). Significant data resourced provided by industry partners in the research, coupled with models and computational resources, will enable better prediction of wind profiles and utilization. In addition, the team will develop dedicated reconfigurable field programmable gate array (FPGA) processors that will be 50 to 500 times faster than general-purpose CPUs for both on-site and central control processing and have small form-factor, low cost and energy efficient to enable agile development under severe outdoor conditions at wind farms.

在航空航天、民用、电气和机械工程领域的集体努力下，风能取得了显着进展。更大的涡轮机被设计和安装，风力发电场现在建在风更间歇和维护设备更难接近的地方。这给确保运行可靠性带来了新的挑战。为了科普这些挑战，沿着微电子技术的快速发展，现代风力发电场配备有大量和各种传感器，包括在涡轮机级的风速计、转速计、加速度计、温度计、应变传感器和功率计，以及在发电场级的风速计、叶片、声纳、温度计、湿度计、压力计等。值得注意的是，所有这些数据目前仅在其各自的领域进行分析/利用。该项目中的大数据挑战包括如何最好地使用时空数据进行风预测，如何使用不同性质的数据（风，功率，负荷等）。和不同来源的数据（物理数据与计算机模拟数据），以计算效率高的方式进行发电评估，最后如何将这三套解决方案集成到一个可靠和高效的计算平台上。拟议的研究和教育活动将通过展示数据科学创新如何使该行业受益来实现风电行业的范式转变。研究所将通过课堂教学、期刊/会议出版物、行业研讨会和数据/软件共享来传播研究成果。暑期实习机会和本科生研究有助于培养下一代劳动力，使其更好地掌握数据科学方法。具有成本效益的风力发电及其普遍采用的关键障碍部分源于风力随机性，严重复杂化了风力发电生产优化和成本降低。风能的长期可行性取决于对其生产可靠性的良好理解，而生产可靠性又受到风力和风力涡轮机发电效率的可预测性的影响。此外，风力涡轮机的生产率包括两个方面：其在其操作期间将风转换成功率的能力和风力涡轮机的可用性。三项相互关联的研究工作将提高风能的可靠性和生产率：（1）时空分析（用于风力预测）（2）条件密度估计（用于风力发电转换评估）;（3）重要性抽样（用于涡轮机可靠性评估和改进）。行业合作伙伴在研究中提供的重要数据资源，加上模型和计算资源，将有助于更好地预测风廓线和利用率。 此外，该团队还将开发专用的可重配置现场可编程门阵列（FPGA）处理器，其速度将比现场和中央控制处理的通用CPU快50至500倍，并且具有小尺寸，低成本和高能效，以便在风力发电场的恶劣室外条件下实现敏捷开发。

项目成果

Evaluation of alternative power production efficiency metrics for offshore wind turbines and farms

• DOI: 10.1016/j.renene.2018.05.050
• 发表时间: 2018-12
• 期刊: Renewable Energy
• 影响因子: 8.7
• 作者: B. Niu;Hoon Hwangbo;L. Zeng;Yu Ding

Dynamic Heterogeneous Voltage Regulation for Systolic Array-Based DNN Accelerators

基于脉动阵列的 DNN 加速器的动态异构电压调节

• DOI: 10.1109/iccd50377.2020.00088
• 发表时间: 2020
• 期刊: 2020 IEEE 38th International Conference on Computer Design (ICCD
• 作者: Chen, Jianhao;Riad, Joseph;Sanchez-Sinencio, Edgar;Li, Peng
• 通讯作者: Li, Peng

Wind Turbine Gearbox Failure Detection Through Cumulative Sum of Multivariate Time Series Data

• DOI: 10.3389/fenrg.2022.904622
• 发表时间: 2022-05
• 期刊: Scientific Reports
• 影响因子: 4.6
• 作者: E. Latiffianti;S. Sheng;Yu Ding

A Scalable FPGA Engine for Parallel Acceleration of Singular Value Decomposition

用于并行加速奇异值分解的可扩展 FPGA 引擎

• DOI: 10.1109/isqed48828.2020.9137055
• 发表时间: 2020
• 期刊: 2020 21st International Symposium on Quality Electronic Design (ISQED
• 作者: Wang, Yu;Lee, Jeong-Jun;Ding, Yu;Li, Peng
• 通讯作者: Li, Peng

Analysis of leading edge protection application on wind turbine performance through energy and power decomposition approaches

通过能量和功率分解方法分析风力涡轮机性能的前沿保护应用

• DOI: 10.1002/we.2722
• 发表时间: 2022
• 期刊: Wind Energy
• 影响因子: 4.1
• 作者: Latiffianti, Effi;Ding, Yu;Sheng, Shawn;Williams, Lindy;Morshedizadeh, Majid;Rodgers, Marianne
• 通讯作者: Rodgers, Marianne