Data mining for power system security assessment-a large scale study

Power system security assessment studies are carried out on a regular basis off-line by expert engineers in order to verify the reliability of a system and take appropriate decisions when necessary. Because of the necessity to open the transmission system to a maximum number transactions to avoid discrimination and/or under-utilization of the system, there is a strong pressure from various actors in order to push the security limits as far as possible, so as to avoid unjustified uplift costs. At the same time, the environment of the transmission system has become less predictable than in the past. Thus, engineers are now faced with more uncertainties and higher pressure to optimize and justify their decisions in the context of security. Consequently, there is also a stronger need for security assessment methodologies and tools which would allow to increase the quality of decision rules designed, while explicitly taking into account a number of uncertainties and helping to objectively justify the decisions. This paper presents a statistical methodology based on a combination of Monte-Carlo sampling of scenarios and data mining to exploit the resulting simulations so as to produce decision rules. The generic methodology consists of three steps: automatic security information database generation; data mining; and use of the synthetic information to take appropriate decisions as concerns security assessment and enhancement, in off-line or on-line mode.