The relationship between the technical condition of components and the reliability of electricity supply  

New knowledge, methods and tools 

To be able to integrate new load demand and generation in an ageing power system, we need to better utilize the components of the power system. Power transformers are examples of such components and are used as examples throughout this blog post since they are expensive and important components in the power system.  

At the same time, we also need to keep reliability of electricity supply at an acceptable level. Reliability of supply is related to the frequency of power supply interruptions and how long they last. 

To utilize and manage components better, power system planners need better information on the actual risk exposure of the power system and how it is affected by different factors. The technical condition of power system components is an important risk-influencing factor. 

In the knowledge-building project VulPro (Risk and Vulnerability Prognosis for power system development and asset management), SINTEF Energy Research has been collaborating with NTNU, Statnett, Landsnet and NVE to give power system planners new knowledge, methods and tools that capture the relationship between component condition and power system reliability. VulPro has been co-funded by the Research Council of Norway. 

 In this blog post we summarize the project’s results, conclusions and recommendations, and we outline some next steps for applying condition- and risk-based methods in power transmission and distribution systems. 

The main research challenges of the VulPro project 

Power system planning includes both long-term system development and mid-term asset management. Examples of decisions within power system development are where and when to reinforce or expand the grid, while mid-term asset management includes decisions about existing assets, such as when to renew or maintain power transformers.  

Traditionally, the risk analyses used to inform mid-term asset management decisions focus on single components in the power system, for example transformers, but do not properly account for their importance for the reliability of supply. Risk analyses for power system development use a more simplified model of the power system that often neglect individual components and their characteristics. They do not account for how a component’s condition influences its probability of failure and how this contributes to the risk with regard to the reliability of supply.  

These two perspectives are illustrated in Figure 1 below. VulPro has contributed to bridging this “gap” between individual components and the power system in analyses of reliability of supply.  

Figure 1: Overall challenge of the VulPro project: Bridging the gap between individual components (left) and the power system (right) in analyses of risk with respect to reliability of supply. 

Transformer wear-out failures due to poor technical condition are very rare, but if a transformer does break down, it takes a long time to replace, and the consequences can be significant. To complement the work on transformers, the PhD study in the project has focused on circuit breakers and on how their technical condition impact reliability of supply. 

The main conclusions and recommendations from the VulPro project 

The project has developed methods for estimating the risk with respect to reliability of supply associated to alternative asset management and system development decisions and how it may evolve over the next years and decades.  

Figure 2 illustrates the main recommendations from the project as a step-wise transition towards better use of risk-based methods: Each step gives system planners more comprehensive information about the risk with respect to reliability of supply, and to a greater extent, captures the power system perspective and not only the component perspective.  

Figure 2: Steps to better use of risk-based methods to support power planning decisions.  

Figure 2 The main conclusions and recommendations from the project are summarized even more briefly as follows: 

1. Component age should not be the primary basis for asset management decisions. 

2. Most components are well maintained and in good condition and have a low probability of wear-out failure.  Data can be used to estimate if the condition deteriorates, and the probability of failure will increase. 

3. Risk-based asset management strategies will likely give considerable lifecycle savings on reinvestment and maintenance costs while keeping the risk to reliability of supply at a satisfactory level. 

4. Risk-based strategies should also consider the uncertainty and vulnerability with respect to long outage durations after component failures. 

5. Mid-term asset management decisions should be made in accordance with the future system development alternatives. Correspondingly, system development plans should consider the needs of the existing assets and mid-term asset management decisions. 

In addition, you can find recommendations for continued data collection and other suggestions for further work at the end of this blog post. 

Step 1: Component age and component condition 

Ageing power system components are a concern for reliability of supply, but as shown also in previous projects, age by itself does not tell us that much about a component’s probability to fail. What matters is how the technical condition of the component has degraded over time. Therefore, component age should not be the primary basis for asset management decisions.  

Historically, there has for most components a weak correlation between component age and component condition, as illustrated in Figure 3. This is partly because degradation is held back by effective preventive maintenance efforts and partly because of preventive replacement (renewal) of components as they age.  

Figure 3. There is a weak correlation between component age and component condition 

A relationship between age and condition, as illustrated in Figure 3, has been substantiated for power transformers using SINTEF’s transformer health index model and for high-voltage circuit breakers in the PhD study in VulPro. 

Step 2: The impact of the condition on component reliability 

Traditionally, reliability of supply analysis in system development studies does not consider that different components have different technical condition. Instead, the analysis usually assumes that all components of the same type have the same average historical failure probability.  

Figure 4 shows how the methods developed in the project gives more information on the failure probabilities for individual transformers. Using for example a transformer health index model we can estimate how the technical condition of a component impacts its reliability. 

Figure 4. Transformers with different technical condition have different probabilities to fail.

For the current fleet of power transformers in Norway, the failure probabilities of the transformers are relatively similar because the transformers are generally in good condition. However, if the technical condition is allowed to deteriorate, the failure probability will increase, and it becomes even more important to account for differences in component condition and how it impacts the reliability of supply.

Step 3: The impact of condition degradation on reliability of supply

The impact a component’s technical condition has on the reliability of supply also depends on its importance, given by its location and role in the power system.

To estimate how condition impacts the reliability of supply, we first had to integrate the component reliability model described above into power system reliability models. You can read more about methods for doing this in this previous blogpost from the project (in Norwegian).

System development and asset management decisions have time horizons of several years or decades. To inform such decisions, we had to develop methods to forecast how the reliability of supply might evolve over these time horizons, for instance measured in MWh energy not supplied expected per year.  Because this gives a measure of the risk to reliability of supply, we will use the shorthand term “risk prognosis”.

These risk prognoses should consider not only the current condition of the components, but also how the condition might deteriorate. We have also modelled the effect that mid-term asset management decisions have on component condition, focusing on renewal of transformers.

To illustrate the use of the risk prognoses, we have compared different asset management strategies that considers information influencing the risk to a lesser or greater extent:

1. An age-based strategy prioritizes which transformers to renew according to transformer age
2. A condition-based strategy prioritizes which transformers to renew according to the technical condition of the transformers
3. A risk-based strategy prioritizes transformer renewal according to both the importance of each transformer for the reliability of the power system and its condition-dependent failure probabilities

For all three strategies we model the same use of resources, or in other words that the use of money and time is the same across the strategies. This is done to be able to compare their performance in terms of reliability of supply in a fair manner.

Figure 5 illustrates how the risk prognoses differ for the three strategies. Because transformers are ageing very slowly, a time horizon of 40 years is chosen for these prognoses. In practice, many other things are changing in the power system over this time horizon. These prognoses thus only show the effect of 1) transformer ageing and 2) transformer renewal, all other things being equal.

Figure 5. Comparison of reliability of supply for different asset management strategies. (CI = Confidence Interval, which is a measure of the uncertainty in the result.)

There is a great deal of uncertainty about how fast each component will degrade. Therefore, it is important to include this uncertainty in the risk prognoses, and that is shown by coloured bands in Figure 5. To consider only the mean values shown by the black curves in the figure may give an incomplete picture of the risk.

A risk-based strategy gave better reliability of supply than the condition- and age-based strategies in this example. This is because the strategy helps in better prioritizing and timing the use of resources. This implies that risk-based strategies can achieve a given level of reliability of supply with lower use of resources than the other strategies.

Step 4: Vulnerability related to long component outage durations

The risk-based asset management strategy (c) above may give lower priority to a transformer with poor condition if its role in the power system is less critical than other transformers. This can be the case for example if it supplies a relatively small share of the load in the system.

However, if this transformer does break down, the consequences may be critical for the customers it supplies. Therefore, the power system becomes vulnerable if the transformer condition deteriorates too much.  Moreover, there is also great uncertainty in the outage duration, which is how long a transformer is out of service after it has failed.

The outage duration depends on the availability of spare transformers, accessibility to the fault site, and when the failure happens during the year. It also depends on technical condition and the type of failure, since wear-out failures can give a longer outage duration. The longer the outage, the more vulnerable the system is to overlapping outages from other components, which may lead to power supply interruptions.

To estimate the outage duration for a transformer and its uncertainty, the project has developed a model that is implemented in open-source Python code and is available at: https://gitlab.sintef.no/power-system-asset-management/t_odm

The risk of an event, such as a transformer failure, is a combination of the probability and the consequence of the event. If the outage duration of the transformer after the failure is longer than expected, then the consequences in terms of power supply interruptions may be greater than expected. Therefore, it is important to take into account the uncertainty in the outage duration.

Figure 6 shows an example of a risk diagram for such an event for the case that the consequence of a transformer failure is considered acceptable if we only consider the expected value. When we use the transformer outage duration model, on the other hand, we also get information about the uncertainty in the outage time and how it affects the uncertainty in potential power supply interruptions. The figure shows an example where potential consequences are uncovered that are unacceptable according to the operational policy or risk policy of the grid owner. This is an example of how vulnerabilities related to long component outage durations can be considered in risk-based methods.

Figure 6. Risk diagram showing the potential consequences of for example a transformer failure.

Step 5. Considering system development as well as component degradation

For the fifth and final step, we consider the following, simplified example of a mid-term asset management decision: When to renew an ageing transformer in a power grid where a new industrial load is seeking to be connected.

Here there are several things that are changing over the time horizon: In addition to 1) transformer ageing and 2) transformer renewal as we considered previously (Figure 5), we also have 3) increasing load demand and 4) grid development. Figure 7 below shows the risk prognosis for two grid development scenarios: In scenario a on top, the grid is reinforced before the new load is connected, whereas in scenario b below, the grid is not reinforced. Scenario b requires that the grid owner accepts a higher risk to reliability of supply. At a later point, the ageing transformer is renewed, which contributes to reducing the risk to reliability of supply.

Figure 7. Risk prognoses illustrating the combined effect of mid-term asset management (here: transformer renewal) and system development (here: connecting new major load, without (above) or with (below) reinforcing the grid first).

For these two cases, the methods in the previous sections were used to illustrate the following points:

• Allowing the load demand in the system to increase without reinforcing the grid first makes the system more vulnerable to condition-dependent (wear-out) transformer failures.
• The grid owner can use the uncertainty information in the risk prognosis. For instance, the grid owner can look at the upper bound of the forecast shown with the lightly coloured band and consider whether it is acceptable to let it stay at such a high level from 2030 to 2035.
• The choices made for system development influence mid-term asset management decisions. In Figure 7, the timing of the transformer’s renewal is the same with and without grid reinforcement. Then using a probabilistic criterion that minimizes the socio-economic costs, we found that we could postpone the renewal of the transformer by around 10 years if the grid is reinforced first. In other words, there is a benefit of considering system development and asset management decisions together.
• The estimate of the duration of possible outages of the transformer greatly influences the optimal timing of its renewal. If one expects a longer outage duration, the transformer should be renewed earlier.

Next steps

The methodologies developed in the project can be implemented in the power transmission and distribution industry. But there are some changes organizations should make: Decision processes within the grid owners must be updated to better consider risk information, and the competence on risk-based and probabilistic approaches should be improved.

The project has demonstrated the value of closer collaboration between different departments within the organization of the grid owner. Participation of grid owner representatives from both system development and mid-term asset management has been important to capture different perspectives and the interdependencies between the different decision processes.

Risk-based approaches have higher data and information processing requirements. VulPro therefore recommends that power grid owners collect and manage relevant component condition and reliability data. Transformer condition data should be collected in a common data base both for operational transformers and scrapped transformers, together with historic load and temperature data.

Through the research work, the project has also identified needs for more knowledge. The following points are key knowledge gaps which we suggest being closed in subsequent research and development efforts:

• The impact of component loading: How does the reliability of components depend on their utilization, that is, how heavily they are loaded and how dynamic the loading patterns are? This is important to know how future loading patterns and higher utilization of the grid may influence the reliability of supply.
• Optimal grid operation and utilization: How to make decisions in grid operation (e.g., allowing overloading of transformers) or mid-term asset management (e.g., more intensive condition monitoring) that optimize the utilization of the existing grid while more grid is being developed?
• Optimal outage planning: With increasing grid utilization, it becomes harder and harder to take components out of service when necessary for maintenance and grid development, and it requires coordination between operational planning, mid-term asset management and system development.
• Optimal emergency preparedness: Can vulnerabilities related to long outage durations be mitigated by planning for better emergency preparedness, for instance by optimizing the spare part strategy?
• Power grid components: The VulPro project has focused on power transformers, and in addition high-voltage circuit breakers has been considered in a PhD study. Degradation models for these components should be further improved for increased accuracy. Furthermore, there is a great need to develop degradation and reliability models for other components such as power cables and overhead power lines, for which few models are available today.
• Valuating reliability of supply and balancing it against other criteria: The results from the VulPro project quantify the reliability of supply, which is an important criterion in power system planning decisions. But more knowledge is needed about the value of reliability of supply in monetary terms, and how it can be balanced against other criteria in practical decision making.

SINTEF will continue to develop and the support the implementation of risk-based methods for power system planning. For instance, some topics will be considered in the newly launched centre for environment-friendly energy research SecurEL.