Computer systems that learn: classification and prediction methods from statistics, neural nets, machine learning, and expert systems
Automatic Learning Techniques in Power Systems
Automatic Learning Techniques in Power Systems
Neural Networks: A Comprehensive Foundation
Neural Networks: A Comprehensive Foundation
Can machine learning be secure?
ASIACCS '06 Proceedings of the 2006 ACM Symposium on Information, computer and communications security
Computational Intelligence Techniques for a Smart Electric Grid of the Future
ISNN '09 Proceedings of the 6th International Symposium on Neural Networks on Advances in Neural Networks
| Hi-index | 0.00 |
Security assessment is a major concern in planning and operating electric power systems. It consists of evaluating the power system's ability to face various contingencies, and proposing ways to counter its main weaknesses when necessary. Contingencies may be external or internal events (for instance, faults subsequent to lightning versus operator-initiated switching sequences) and may consist of small/slow or large/fast disturbances (for example, random behavior of the demand pattern versus generator or line tripping). Usually, numerical (for example, time-domain) simulation of the corresponding scenario assesses the effect of a contingency on a power system in a given state. However, the nonlinear nature of the physical phenomena and the growing complexity of real-life power systems make security assessment difficult. For example, monitoring a power system every day calls for fast analysis, sensitivity analysis to identify the salient parameters driving the phenomena, and suggestions on how to act on the system so as to increase its level of security. On the other hand, increasing economic and environmental pressure make the conflicting aspects of security and economy even more challenging. To meet these challenges, we need methods different from the standard time-domain simulation approaches. The author describes ongoing research and development of machine learning and other automatic-learning techniques and their adaptation to the specific needs of power-system security assessment. In particular, the author describes a framework that integrates several of these techniques so that users can extract relevant information tailored to their decision-making needs. Among the many other potential applications of automatic learning in power systems, security assessment is probably the most needed and versatile.