CAREER: Scalable Learning and Adaptation with Intelligent Techniques and Neural Networks for Reconfiguration and Survivability of Complex Systems

近年来，智能技术和适应性批判性设计受到越来越多的关注。电网及其部件等复杂系统的动态随机优化(DSO)问题可以表示为某一参数的最小化和/或极大化。电网面临着放松管制和我们的数字经济对高质量和可靠电力的需求增加，再加上与其他关键基础设施的相互依存，它的压力越来越大。智能系统技术将在DSO的实施中发挥重要作用，在不严重降低可靠性和安全性的情况下，提高网络效率，消除拥塞问题。该项目旨在探索电网动态优化的方法，作为先进的类脑随机辨识器和控制器的试验台，将增进对动态随机系统优化的认识和理解。针对大规模复杂系统，设计了一种新的局部和全局动态随机优化策略。电网等大型复杂系统目前存在的运行安全裕度将被最小化，从而允许最大限度地利用现有资源，提高系统可靠性和安全性，并对整个系统的设备进行优化设置。能够进行动态随机优化是当今的梦想。这一提议是使用基于近似动态规划的学习和适应的类脑系统、高级神经网络和复杂系统上的其他智能技术将这一梦想变成现实的第一步。此外，系统的生存性和可用性将通过提高数字硬件的可靠性和容错性来提高，其中关键算法是使用进化和智能技术来实现的。容错设计对不可预见性意味着健壮性、安全性和安全性。该项目还将包括教育推广和国际合作的一个重要组成部分，包括通过美国与尼日利亚以及美国与巴西之间的教师和学生交流进行智力交流。

Recently, intelligent techniques and adaptive critic designs have received increasing attention. The dynamic stochastic optimization (DSO) of complex systems such as the electric power grid and its parts can be formulated as minimization and/or maximization of certain quantities. The electric power grid is faced with deregulation and an increased demand for high-quality and reliable electricity for our digital economy, and coupled with interdependencies with other critical infrastructures, it is becoming more and more stressed. Intelligent systems technology will play an important role in carrying out DSO to improve the network efficiency and eliminate congestion problems without seriously diminishing reliability and security. This project proposes to investigate ways in which the power grid can be dynamically optimized, as a testbed for advanced brain-like stochastic identifiers and controllers.This project will advance knowledge and understanding on how to carry out optimization of a dynamic stochastic system. A novel local and global dynamic stochastic optimization strategy for a large scale complex system will be designed. The operating safety margins that currently exist on the large complex systems such as the electric power grid will be minimized, thus, allowing maximum utilization of existing resources with increased system reliability and security with optimal settings on devices throughout the entire system. The capability of carrying out dynamic stochastic optimization is the dream of today. This proposal is a first step in unfolding this dream to reality using brain-like systems with learning and adaptation based on approximate dynamic programming, advanced neural networks and other intelligent techniques on complex systems. In addition, system survivability and availability will be increased by improving reliability and fault tolerance of digital hardware, where the critical algorithms are implemented, using evolution and intelligent techniques. Fault tolerant designs to the unpredictable means robustness, security and safety. The project will also include a major component of educational outreach and of international collaboration including intellectual exchange via faculty and student exchanges between the U.S. and Nigeria, and US and Brazil.