Power System Control Under Cascading Failures - Sun, Kai; Hou, Yunhe; Sun, Wei; Qi, Junjian (2024)

Leseprobe

1
Introduction
1.1 Importance of Modeling and Understanding Cascading Failures
1.1.1 Cascading Failures

Cascading failures can happen in many different systems, such as in electric power systems [1-7], the Internet [8], the road system [9], and in social and economic systems [10]. These low-probability high-impact events can produce significant economic and social losses.

In electric power grids, cascading blackouts are complicated sequences of dependent outages that could bring about tremendous economic and social losses. Large-scale cascading blackouts have substantial risk and pose great challenges in simulation, analysis, and mitigation. It is important to study the mechanisms of cascading failures so that the risk of large-scale blackouts may be better quantified and mitigated. Cascading blackouts are usually considered rare events, but they are not that uncommon. The frequency of these high-impact events is not as low as expected. The following is a subset of the very famous large-scale blackouts around the world.
1965 Northeast blackout: There was a significant disruption in the supply of electricity on November 9, 1965, affecting parts of Ontario in Canada and Connecticut, Massachusetts, New Hampshire, New Jersey, New York, Rhode Island, Pennsylvania, and Vermont in the United States. Over 30 million people and 80,000 square miles were left without electricity for up to 13âhours [11].
1996 Western North America blackouts: A disturbance occurred on July 2, 1996, which ultimately resulted in the Western Systems Coordinating Council () system separating into five islands and in electric service interruptions to over two million customers. Electric service was restored to most customers within 30âminutes, except on the Idaho Power Company () system, a portion of the Public Service Company of Colorado (), and the Platte River Power Authority () systems in Colorado, where some customers were out of service for up to 6 hours [12]. The first significant event was a single phase-to-ground fault on the 345-kV Jim Bridger-Kinport line due to a flashover (arc) when the conductor sagged close to a tree. On July 3, a similar blackout occurred, also initiated by the tree flashover of the 345âkV Jim Bridger-Kinport line.
2003 U.S.-Canadian blackout: A widespread power outage occurred throughout parts of the northeastern and midwestern United States and the Canadian province of Ontario on August 14, 2003, affecting an estimated 10 million people in Ontario and 45 million people in eight U.S. states [13]. The initiating events were the out-of-service of a generating plant in Eastlake, Ohio, and the following tripping of several transmission lines due to tree flashover. Key factors include inoperative state estimator due to incorrect telemetry data and the failure of the alarm system at FirstEnergy s control room.
2003 Italy blackout: There was a serious power outage that affected all of Italy - except the islands of Sardinia and Elba - for 12âhours and part of Switzerland near Geneva for 3âhours on September 28, 2003. It was the largest blackout in the series of blackouts in 2003, affecting a total of 56 million people [14]. The initiating event was the tripping of a major tie line from Switzerland to Italy due to tree flashover. Then a second 380-kV line also tripped on the same border (Italy-Switzerland) due to tree contact. The resulting power deficit in Italy caused Italy to lose synchronism with the rest of Europe, and the lines on the interface between France and Italy were tripped by distance relays. The same happened for the 220-kV interconnection between Italy and Austria. Subsequently, the final 380-kV corridor between Italy and Slovenia became overloaded and it too was tripped. Due to a significant amount of power shortage, the frequency in the Italian system started to fall. The frequency decay was not controlled adequately to stop generation from tripping due to underfrequency. Thus, over the course of several minutes, the entire Italian system collapsed, causing a nationwide blackout [15].
2012 Indian blackout: On July 30 and 31, 2012, there was a major blackout in India that affected over 600 million people. On July 30, nearly the entire north region covering eight states was affected, with a loss of 38â000âMW of load. On July 31, 48â000âMW of load was shed, affecting 21 states. These major failures in the synchronously operating North-East-Northeast-West grid were initiated by overloadin of an interregional tie line on both days [16-18].
2015 Ukrainian blackout: On December 23, 2015, the Ukrainian Kyivoblenergo, a regional electricity distribution company, reported service outages to customers [19]. The outages were due to a third party s illegal entry into the company s computer and supervisory control and data acquisition () systems: Starting at approximately 3:35âp.m. local time, seven 110-kV and 23 35-kV substations were disconnected for 3 hours. Later statements indicated that the cyber-attack impacted additional portions of the distribution grid and forced operators to switch to manual mode. The event was elaborated on by the Ukrainian news media, who conducted interviews and determined that a foreign attacker remotely controlled the distribution management system. The outages were originally thought to have affected approximately 80,000 customers, based on the Kyivoblenergo s update to customers. However, later it was revealed that three different distribution companies were attacked, resulting in several outages that caused approximately 225,000 customers to lose power across various areas.
2016 Southern California disturbance: On August 16, 2016, the Blue Cut fire began in the Cajon Pass and quickly moved toward an important transmission corridor that is composed of three 500-kV lines owned by Southern California Edison () and two 287-kV lines owned by Los Angeles Department of Water and Power () [20]. The transmission system experienced 13 500-kV line faults, and the system experienced two 287-kV faults because of the fire. Four of these fault events resulted in the loss of a significant amount of solar photovoltaic () generation. The most significant event related to the solar generation loss occurred at 11:45âa.m. Pacific Time and resulted in the loss of nearly 1200âMW.
2016 South Australia (SA) blackout: On September 28, 2016 there was a widespread power outage in SA power grid which caused around 850,000 customers to lose their power supply [21]. Before the blackout the total load including loss in the SA power grid was 1826 MW, among which around 883 MW was supplied by wind generation, corresponding to a very high renewable penetration [22]. Late in the afternoon a severe storm hit SA and damaged several remote transmission towers. The SA grid subsequently lost around 52% of wind generation within a few minutes. This deficit had to be compensated by the power import from the neighboring state, Victoria, through the Heywood AC interconnection. The significantly increased power flow was beyond the capability of the interconnection. Ultimately the SA system was separated from the rest of the system before it collapsed [22].

Some of the past cascading blackouts share similarities. For example, the two significant outages in the western North America in 1996 [12], the U.S.-Canadian blackout on August 14, 2003 [13], and the outage in Italy on September 28, 2003 [14], all had tree contact with transmission lines [23]. Modeling and understanding these common features will help prevent future cascading blackouts that might be initiated by the same reason. At the same time, each blackout has its own unique features due to the characteristics of the particular system, which makes the modeling and understanding of cascading failures challenging.
1.1.2 Challenges in Modeling and Understanding Cascading Failures

The modeling and understanding of cascading failures, or in particular cascading blackouts, can be very challenging in the following aspects:
Size of the system: The size of the interconnected power system can be very large. For example, in the United States utility companies build power system models, which are then used to create the North American Electric Reliability Corporation () interconnection-wide models, with over 50â000 buses. Modeling and understanding the possible ways that such a big system fails can be really challenging.
Limited computational power: The computational power is constantly improving as technologies for both hardware and software advance. However, it is still very limited. Although Nâââ1 contingency analysis is usually achievable, even only Nâââ2 contingency analysis for a system with thousands of components can lead to formidable computational burden [24].
Mechanisms in cascading blackouts: There can be many mechanisms during a cascading blackout, which can include thermal dynamics of the...