Neural Networks for Estimating and Compensating the Nonlinear Characteristics of Nonstationary Complex Systems

用于估计和补偿非平稳复杂系统非线性特性的神经网络

• 批准号: 0601521
• 负责人: Ronald Harley
• 金额: $ 24万
• 依托单位: Georgia Tech Research Corporation
• 依托单位国家: 美国
• 项目类别: Standard Grant
• 财政年份: 2006
• 资助国家: 美国
• 起止时间: 2006-05-01 至 2010-04-30
• 项目状态: 已结题
• 来源: https://www.nsf.gov/awardsearch/showAward?AWD_ID=0601521&HistoricalAwards=false
• 关键词: Neural; Networks; Estimating; Compensating; Nonlinear

NEURAL NETWORKS FOR ESTIMATING AND COMPENSATING THE NONLINEAR CHARACTERISTICS OF NONSTATIONARY COMPLEX SYSTEMSABSTRACTScope and Intellectual Merit. The objective of this research is to find a method of accurately quantifying the distorted currents and voltages created by certain devices in power networks. Distortion causes electromagnetic inference with communication and the fast growing digital world, light flicker, overheating of electric machines and transformers and increased losses in transmission lines. For years utilities and customers have argued about who causes the distortion. Existing measurement techniques can lead to errors of up to 40%. The approach is to use Echo State Networks and Simultaneous Recurrent Neural Networks with super fast learning algorithms (biological inspired algorithms such as particle swarm optimization), and other computational intelligence algorithms, to accurately measure the distortion by monitoring only voltage and current without the need for added transducers. Such fast and powerful neural networks could also be used for closed loop control of the offending nonlinear devices to mitigate the distortion.Broader Benefits. The economic impact of applying brain-like techniques to monitor and control physical processes is significant. Reduced power losses mean savings and more useful power over the same lines. More secure and reliable power systems of high quality are of national interest. Moreover, reduced electromagnetic interference promotes a cleaner more reliable telecommunications and digital environment. Fast intelligent nonlinear controllers will also benefit other real-world high-speed closed loop controlled nonlinear non-stationary processes. There exists a talent shortage in the US in the application of intelligent systems, and the project will train a new generation of professionals, and educators, underrepresented minorities and undergraduates in the multiple fields of the project

估计和补偿非平稳复杂系统非线性特性的神经网络及其智能价值这项研究的目的是找到一种准确量化电网中某些设备产生的失真电流和电压的方法。失真导致通信和快速增长的数字世界的电磁干扰、灯光闪烁、电机和变压器过热，以及输电线路损耗增加。多年来，公用事业公司和客户一直在争论是谁造成了扭曲。现有的测量技术可能会导致高达40%的误差。该方法是使用回声状态网络和同时递归神经网络与超快学习算法(粒子群优化等生物启发算法)和其他计算智能算法相结合，通过仅监测电压和电流来精确测量失真，而不需要增加换能器。这种快速而强大的神经网络还可以用于对违规的非线性设备进行闭环控制，以减轻失真。应用类似大脑的技术来监测和控制物理过程的经济影响是显著的。更低的功率损耗意味着在相同的线路上节省和更多有用的功率。更安全、更可靠的高质量电力系统符合国家利益。此外，电磁干扰的减少促进了更清洁、更可靠的电信和数字环境。快速智能非线性控制器也将有益于其他真实世界中高速闭环控制的非线性非平稳过程。美国在智能系统应用方面存在人才短缺，该项目将在该项目的多个领域培养新一代专业人员、教育工作者、未被充分代表的少数民族和本科生