Major blackouts are usually caused by cascading contingencies, such as a short circuit, an overloaded component, and a generator outage, with complicated interactions. The vulnerability of the system to (in itself) low-probability incidents that expand to a cascading outage (which is also called the domino effect) increases when the system is already stressed by other causes, such as congested transmission corridors when there is a bulk exchange of power between parts of the system. Quite often, a cascading outage is initiated by forces of nature or by weather conditions: (thunder) storms, extreme temperatures, geomagnetic storms, forest fires, and so forth. The sequence of events leading to a blackout are usually diverse, but the result is always the same: an interruption of the power supply for a certain period of time.

Two major blackouts in 2003 show what can cause such a disastrous event and on what time scale it takes place. The first  blackout happend in the northeastern part of the United States and Canada on August 14, 2003; the blackout affected approximately 50 million people, and it took more than 24 hours to restore the power supply in New York City and other areas. The second took place in Italy on September 28, 2003. The blackout affected approximately 57 million people, and it took 5–9 hours to restore the power to Rome and the major cities of the country.

Blackout in Northeastern United States and Canada (August 14, 2003)

The blackout in Northeastern United States and Canada has been investigated and documented by the North American Electric Reliability Council [1] and the US–Canada Power System Outage Task Force [2]. Here follows a short description of the sequence of events that led to the blackout in Northeastern United States and Canada (see Figure 1 ).On August 14, 2003, there were voltage problems in the border area between the United States and Canada. The scenario that led to the actual blackout in New York and wide surroundings at 4.13 p.m. started around noon. Shortly after 12 o’clock, a 375 MW unit in the Conesville power station in mid-Ohio was disconnected from the grid, followed by 785 MW of the Greenwood power station in northern Detroit (Michigan) 1 hour 10 minutes later and 597 MW of the Eastlake power station in northern Ohio 18 minutes later. The loss of these three generating units caused a change in the power flow in the power pool around the great lakes, the so-called Eastern Interconnection. At 2.02 p.m., an important 345 kV transmission link from the southwest to the north of Ohio was taken out of service. There was a forest fire close to the transmission lines, and there was fear that the heat would ionize the air surrounding the lines and would cause a short circuit. Between 3.00 p.m. and a 3.45 p.m., three other 345 kV links were taken out of service, and an important connection in the transport capacity between east and north Ohio had at that moment disappeared. The power flow divided up over the remaining connections, such as connections at the 138-kV level. These connections got overloaded. Because of the large voltage drop, 600 MW industrial consumption was disconnected in Ohio, as well as consumers at the 138 and 69 kV network.

Because of the loss of the 345kV connections Canton Central–Tidd and Sammis–Star, the power could flow to the north of Ohio by means of three routes only (see Figure 8.8, I). The north of Ohio, which normally provided the east of Michigan with energy, became a weak point; the large industrial town Detroit depended on southeast Michigan for its power supply. Subsequently, the connections Galion–Ohio, Central–Muskingum (Figure 8.8, II), and East Lima–Fostoria Central (Figure 8.8, III), all 345kV lines, were taken out of service, and the link between north Ohio and the east of Michigan was formed mainly by two 345 kV connections at the south side of Lake Erie. The power flows from Indiana and the lines in southwest Michigan in the direction of east Michigan and north Ohio increased but were insufficient to meet the demand, so that the voltage in north Ohio dropped.

Shortly after that, a number of large power plants were disconnected from the grid in north Ohio and west Michigan. It was 4.10 p.m. The increased power flows overloaded the remaining connections in service, and they were taken out of service. The problem got worse, and when, at a certain moment in time, the 345-kV line Perry–Ashtabula–Erie West was taken out of service, the area covering east Michigan and north Ohio had hardly any production left, and the voltage collapsed.

Subsequently, the line at the south side of Lake Erie was taken out of service, and the power flow changed its direction and flows, with a wide curve, from Pennsylvania to New York to Ontario and Michigan, around the north side of Lake Erie. This also drew Pennsylvania, New York, Ontario, and Quebec into the sequence of events. Two 345kV and two 220kV lines between Pennsylvania and New York were disconnected within 4 seconds of each other due to the suddenly increased power flows. As a result, Pennsylvania and New York were no longer interconnected in a direct way, and large power flows appeared on the New York–New Jersey connection. At the same time, more generation was lost in north Ohio, and also the Fostoria Central–Galion 345 kV connection was idle. The disconnection of the Beaver–Davis–Besse 345-kV line cut Cleveland (Ohio) off from the Eastern Interconnection. Michigan was still connected to Ontario in Canada, but two 230 kV connections became inactive, and Ontario was connected to Manitoba and Minnesota only.

Around that time, the last link between the Eastern Interconnection and the area of New Jersey, the 500-kV connection Branchburg–Ramapo, went out of operation. As a consequence, the supply area around Greater New York was split in two parts: New England (except for southwest Connecticut) and the Maritimes. Large areas were islanding now and tried to reestablish the power balance. Also, Ontario tried to balance the power and shed 2500 MW of load. The direct interconnection via the Niagara Falls between Ontario and New York was taken out of service, and the power flows were now heading for New York from Ontario via Quebec through the 765 kV lines. The recovery of the connection between Ontario and New York failed, and 4500 MW of generation was disconnected in Ontario.

Now, the power supply in the largest part of Ontario was interrupted. Problems in maintaining the power balance in the islanded parts of New York had as a result that the power supply also failed there. A major part of the Eastern Interconnection, the area in the United States around the great lakes, was now without electricity by which time it was 4.13 p.m.


  1. NERC: Technical Analysis of the August 14, 2003, Blackout: What Happened, Why, and What Did We Learn? , July 13, 2004, North American Electric Reliability Council.
  2. U.S.–Canada Power System Outage Task Force: Final Report on the August 14, 2003 Blackout in the United States and Canada: Causes and Recommendations, April 2004, U.S.–Canada Power System Outage Task Force.

Blackout in Italy (September 28, 2003)

The blackout in Italy has been investigated and documented by an investigation committee of the UCTE [1]. There follows a short description of the sequence of events that led to the blackout of the entire Italian peninsula. An overview of the Italian cross-border connections is shown in Figure 2.

The blackout in Italy was initiated by the loss of the Swiss 380 kV transmission line connection between Mettlen and Lavorgo. This high-voltage overhead line was 85% loaded at that time. The line was taken out of service after a short circuit between the line conductors and the branches of a tree that came too close to the line.

Such a short circuit could disappear spontaneously when the arc (a lightning-like conducting path, in this case between the conductor and a branch of the tree) extinguished by itself because of the cooling by the sur- rounding air. Probably this was not the case here, as the automatic reclosure of the high-voltage transmission line failed a number of times. Also, by means of a manual reclosure command, seven minutes after the disturbance, the line could not be brought back into service. The phase angle between the voltages at the ends of the line was too large: 42∘, whereas 30∘ is the maximum allowed phase difference between two points that are to be connected. When this happened, the Italian electricity import from Switzerland was 540 MW more than scheduled and the import from France 440 MW less than scheduled. At the same time, the remaining transmission links took care of the power transport, as dictated by the laws of physics; the major part of the transport capacity was taken over by the lines nearby. This resulted in a 110% loading of the Swiss 80kV link Sils–Soazza, the so-called San Bernardino link. This overload is permitted, but may not last longer than 15 minutes.

Ten minutes after the start of the problems with the line Mettlen–Lavorgo, a discussion by phone was held between the ETRANS control center in Laufenburg, which is located at the border between Germany and Switzerland, and the GRTN control center in Rome, Italy. GRTN was asked to take measures in the Italian area in order to reduce the overload on the Swiss transmission lines and bring back a safe system operation. The measure that had to be taken was in fact the reduction of the Italian import by 300 MW, because that was the surplus amount of imported power at the time of the disturbance. The 300 MW import reduction was a fact 10 minutes after the telephone conversation, and the power balance in Italy was restored again. Also the Swiss took their measures, and, together with the Italian import reduction of 300 MW, the transmission lines in Switzerland were normally loaded again.

But four minutes later, another short circuit occurred between a treetop and a transmission line conductor – this time on the circuit Sils–Soazza, the connection that had been overloaded for 10minutes. Because of the overload, the aluminum conductors were heated and the line sag became larger. As a result, the line conductors came closer to the tops of the trees, as illustrated in Figure 8.10. The connection Sils–Soazza also had to be taken out of service. The loss of these two interconnections resulted in dangerous overloads on the other system components and connections in the system (4 seconds after the disappearance of the line Sils–Soazza, the 220-kV line from

Airolo to Mettlen in Switzerland was overloaded and taken out of service), and therefore, 12seconds after putting the line Sils–Soazza idle, the Italian network was isolated from the rest of Europe in order to reduce the affected area. During these 12 seconds, large power fluctuations in combination with severe transient voltage instabilities occurred.

The voltage drop in the northern part of Italy caused a number of power plants to disconnect automatically from the grid. The power balance was disturbed again, and, because Italy was now isolated from the rest of Europe, the frequency dropped to 49 (see Figure 3). The primary frequency control halted the drop in the frequency by load shedding and by stopping the pumps that elevate river water to reservoirs (for energy storage). But this was already too late: the protection of turbines, underfrequency relays, and temperature relays operated. The shedding of even more load did not help anymore, and, two and a half minute after the isolation of Italy from the European network, the frequency reached the absolute lower limit of 47.5 Hz and the lights went out throughout Italy.

A description of recent major blackouts worldwide.


  1. UCTE: Final Report of the Investigation Committee on the 28 September 2003 Blackout in Italy, April 2004, Union for the Coordination of Transmission of Electricity.4Reliability and Risk Assessment for Electric Power Grids to prevent Blackouts

Reliability and Risk Assessment for Electric Power Grids to prevent Blackouts

by Dr. Simon Tindemans Delft University of Technology

On January 8th, 2021, the European electricity grid was the closest it had been to a large-scale blackout in 15 years. The fact that most people didn’t notice anything was due to a range of risk-avoidance measures that were designed for precisely such an unlikely scenario.

To be more precise, the disturbance affected the synchronous area of continental Europe, which spans from Portugal to Denmark and all the way East to Poland and Turkey. The electricity grid in this area is tightly integrated and operates with a single shared frequency of 50 Hz: the AC voltage signal you receive in the power socket is the same wherever you go in this region, which necessitates tight coordination to keep local disturbances from spinning out of control.

On January 8th, there were large power flows from the South-East to the North-West of Europe, leading to heavy loading on high-voltage transmission equipment. In the early afternoon, overcurrent protection was triggered in a substation in Ernestinovo in Croatia, which caused the East-West power flows in that substation to be diverted to neighbouring power lines, causing further overloads and disconnections. Within a minute, this cascade resulted in the complete separation of the system into two parts (see Figure 5).

Prior to the system separation, 6.3GW of power was flowing between the two regions – equal to one third of the Dutch power consumption on a cold winter night. The separation left the North-West area with a sudden deficit of power and the South-West with an equally large surplus. The imbalances caused immediate large frequency swings, shown in Figure 6, which risked triggering large-scale system instability and perhaps a widespread blackout. Fortunately, automatic response services proved able to rapidly restore the system balance in each of the two subsystems, and they were reconnected approximately one hour after the initial event (evidence by the coalescence of the frequency traces in Figure 2).

This recent event is an exemplary case for power system reliability assessment at the very large scale: accurately gauging the necessary amount of response services is essential to prevent blackouts. But most disturbances people encounter happen at the distribution network level, close to home? In 2019, Dutch electricity customers were disconnected from the grid for an average of 20 minutes, overwhelmingly caused by disruptions in the distribution network.

Access to electrical power is essential in today’s society, and power system reliability analysis is the study of the ability to supply this power to end users, both local and at the bulk scale. As a branch of engineering, it combines analysis (what are the weaknesses) with a focus on design (how can these weaknesses be addressed). It addresses a large variety of questions, including: “Is it wise to spin up the coal-fired power plant to compensate for a forecast dip in wind speeds?”, “Is it worth investing in redundant transformers to reduce customer outages?”, “Do we need to invest in large-scale storage systems to deal with wind-still dark winter days?”

At the core of all such questions lies a risk assessment along the following lines. Firstly, consider all sources of uncertainty, identifying events that could interfere with the successful operation of the power system, such as lightning and squirrel activity(!). Secondly, consider the likelihood of occurrence of these events and the impact they would have if they were to occur, combining them (qualitatively or quantitatively) as risk = probability x impact. And finally, determine whether and how this risk can be mitigated.

Current challenges

As one can imagine, practical limitations in terms of data, modelling and computation ability mean that each of these steps is approximate. Historically, risk assessments were often qualitative in nature, for example by requiring a certain degree of asset redundancy. However, advances in sensing and computation mean that the industry is evolving towards quantitative risk assessment methods. At the forefront of this development, research is done by electrical engineers in close collaboration with statisticians and computer scientists. A selection of active research areas is listed below:

Weather and climate modelling

The importance of accurate weather and climate models is increasing. Besides direct impact of weather on network infrastructure (lightning, storms, etc.), it also affects the output of renewable generators. Moreover, the common practice of extrapolating the weather of the last few decades to the future is challenged by climate change. Finally, and perhaps surprisingly, space weather is gaining recognition as a subject of interest, because geomagnetic storms can induce significant currents in long-distance transmission lines.

Reasonable capacity payments

The rollout of large-scale renewable generation means that conventional gas- or coal-fired generators are slowly being phased out. Not unreasonably, this raises concerns about the security of supply: what can we do to ensure the lights stay on in the event of a Dunkelflaute, an extended period of darkness and little wind? In response, a number of countries are rolling out capacity mechanisms, where generators or batteries receive payments to remain “on standby” for such cases. But do we really need these schemes, and if we do, how much capacity should be purchased at what cost?

AI and machine learning

Last but not least, recent developments in AI and machine learning will fundamentally change the process of reliability analysis and risk assessment. In the short term, machine learning tools can improve the speed and accuracy of assessments that are currently done by system operators and planners, allowing them to tackle problems of previously unsurmountable complexity. In the future, hybrid intelligence will see AI agents working alongside system operators, learning from these operators and suggesting actions to them.