Resilience of interdependent critical infrastructure

Network of networks approach to power system resilience

Resilience includes the ability of a system to either prevent or minimize the effects of disruptions that arise, and to rapidly recover from or adapt to changing conditions associated with emerging threats. However, resilience assessment of power systems alone would result in an isolated methodology treating risks arising from the rest of the infrastructure as externalities. This has the potential to lead to under-preparedness and greater impact from adverse events. A “network of networks” approach would internalise all the relevant fault modes and clearly map the cascading effects and interactions across multiple infrastructure networks. Hence, there is a need for the industry to define a framework and methodology to assess and improve the resilience of interdependent critical infrastructure.

Sources of disruption

The interdependent and multi domain aspects of power system components are often observed during resilience events or high impact low frequency (HILF) events. Each type of disruption has its own effects on various critical infrastructures and this greatly impacts recovery measures. There are several dimensions to consider:

• The effects of physical damage from hurricanes, earthquakes, floods, and fires differ greatly, both in geographical extent, the type of damage, and the required recovery actions.
• Some disasters are confined to one type of infrastructure, but it is very common for disruptions to electricity blackouts to degrade other infrastructure systems, such as water, communications, fuel supply, transportation, and offices and homes.
• The geographic extent of the disruption is also important. It is much more difficult to recover from a widespread damage / outage than a local one.
• Predictability and speed of how hazards unfold can also differ greatly. Some hazards, such as earthquakes are deemed as fast and unpredictable, but other climate related hazards can be slower, anticipated and predicted.       

Unified resilience metric – from reliability to resilience across networks

Most of the critical infrastructure networks have some measure of assessing reliability of service. When it comes to resilience of infrastructure networks, there are generally two different assessment approaches. Calculated metrics from these two approaches are slightly different.

1. One is to assess the resilience of networks based on data associated with the past resilience events. For past events that have already happened, a possible metric might be Total Loss of Load.
2. The second one is a predictive approach, typically considered in reliability studies. For predictive calculation (usually in planning) Expected Loss of Load / Expected Energy Not Supplied, or Loss of Load Probability (LOLP) are more appropriate.

Combined resilience models require definition of figures of merit, which are individual “infrastructural qualities”, i.e. resilience metrics.

The C4.47 Working Group is expanding on this and came up with a resilience definition. There is a need to undertake a similar approach in the other infrastructure networks and come up with resilience metrics for each individual network. The next step would then be the unification of these resilience assessment methodologies, as a first step towards a unified resilience metric that encompasses all the relevant interdependencies. The limitations of metrics are well-understood but, given the complexity and scale of all the infrastructure systems, metrics and indicators can still provide added value.

The definition is generalised further, in order for it to encompass the different infrastructures that interact with electrical power systems. Commonalities can be observed. Ultimately, the resilience of any infrastructure network depends on the capacity of the system to:

• Absorb a disturbance
• Cope with a disturbance
• Restore normal operation after a disturbance
• Adapt itself, learning from a disturbance

This paper looked at the actions and systems relevant to each type of infrastructure, for electricity, gas, water, telecommunications / cyber and transport. This was transposed against each resilience stage, as defined by the CIGRE C4.47 resilience definition, and the following common themes can be seen at each stage:

• Pre-disturbance (anticipation and preparation) – threat scenarios modelling
• Disturbance progress (absorption) – response and reserve mechanisms, contingency measures, coping capacity
• Post-disturbance (sustainment of critical system operations) – core service provision, such as critical loads, or wholesome water
• Restoration (rapid recovery) – repair and maintenance, removal of temporary measures
• Post-restoration (adaptation) – lessons learnt, investment re-evaluation

The key issue in integrating the resilience metrics is to find a method to “normalise” them, so that they become comparable. This can be done by means of:

1. an overall generic infrastructural quality
2. individual infrastructural qualities, weighted and reduced to a single factor

There is also a need to devise clear links between network operation and these metrics. A common theme can be seen in these resilience assessment approaches, which is the structure and interactions of the network. All the infrastructure systems considered in this paper behave as networks / graphs, and can be analysed as such. Candidate graph-based indicators include:

• centrality
• extended betweenness
• node-to-node resilience

While the network of networks approach to assessing potential damage and failures has been well researched, elucidating the recovery phases of resilience remains a key challenge.

Stress tests are also seen by the UK National Infrastructure Commission as the main method of measuring infrastructure resilience performance across water, energy, transport and other infrastructure systems. A meaningful stress test model that involves multiple infrastructures would consist of a comprehensive list of scenarios that sufficiently cover the state space of the multi-network operation, while assessing the resilience metrics for each scenario.

Finally, instead of modelling an infrastructure with the physical network of links and nodes, one can also model how the infrastructure components interact in a directed network called an influence graph that is different than the physical network. For example, in an electric grid, if any particular transmission line outages, then the influence graph describes the probabilities of which other lines outage next in cascades of outages. The influence graph captures the statistics of how outages spread, and it can be created from simulated or actual data of many sequences of components (such as transmission lines) outaging. This extends naturally to interdependent infrastructures if one can model the probability of a particular component failing in one infrastructure causing outages of particular components in the other infrastructure, such as those in the same location. The combined influence graph would then capture the statistics of cascading outages in and between both infrastructures and could be simulated to produce many representative samples of cascades across both infrastructures.

Conclusion

This paper discussed the types and frameworks of interdependencies between diverse but interacting infrastructure systems. A breakdown of the stages and sources of disruption has been provided, and potential metrics and high-level assessment methodologies have been discussed.

In conclusion, the overarching pattern seems to be that joint resilience metrics of interdependent networks must be derived from a generalised point of view, incorporating aspects of the structure of the “network of networks”, as well as the interactions between components. Detailed operational parameters can be incorporated in the underlying model of the individual metrics. When it comes to the combined metric, though, a normalised or weighted approach is the most easily achievable.

---

• Contact Author: S. SKARVELIS-KAZAKOS

Report available on e-cigre

Free for members
30€ for non-members

WGR_320_1