Risk Assessment

The energy sector including its complex and interdependent technical systems is both a critical infrastructure and key resource for the functioning of today’s society and economy. Therefore, the comparative assessment of accident risks is a central component in a comprehensive evaluation of energy security aspects and sustainability performance associated with our current and future energy system. Accidents in the energy sector often affect people’s health and property, the supply of economic goods and services, and degrade ecosystems and their functions. Thus, the protection of critical infrastructure facilities in the energy sector is of paramount importance because a sufficient and continuous energy supply forms the backbone of our modern society and many of its products, which are relying on interrelated and interdependent information and communication technologies. As a consequence, the interest and demand for more and better data on the assessment of severe accidents has considerably risen because they are the basis for improved risk management and informed decision-making to achieve a safe, secure and sustainable energy supply.

Method

Figure 1 below gives an overview of PSI’s comprehensive and integrated framework for comparative risk assessment. The so-called Energy-related Severe Accident Database (ENSAD) is the core component of this activity because assessments build upon historical experience whenever possible. for example for fossil energy chains and hydropower. For fossil energy chains and hydropower the extensive historical experience available in ENSAD is used, whereas for nuclear a simplified probabilistic safety assessment (PSA) is applied, and evaluations of new renewables are based on a hybrid approach combining available data, modeling, and expert judgment. ENSAD provides a comprehensive, worldwide coverage of energy-related accidents. In the data collection process a multitude of primary information sources are surveyed, and their contents are verified, harmonized and merged. In this way a substantial higher degree of completeness can be achieved compared to databases relying on just one or few sources. The initial scope of the database was limited to technological accidents, but more recently also accidents triggered by natural hazards were included. Intentional attacks on energy infrastructures including terrorist threat have been collected in a separate database. For this purpose, the Energy Infrastructure Attack Database (EIAD) was developed, which is a collaboration between PSI and the Center for Security Studies at ETH Zürich, building upon PSI’s experience with ENSAD.

Risk Fig1.png

Figure 1: Comprehensive framework for comparative risk assessment in the energy sector. Source: Burgherr & Hirschberg (2014).

ENSAD takes a full-chain approach because accidents can occur at all stages of an energy chain and not only at the actual power generation step. In ENSAD, data on all energy-related accidents is collected and classified into energy chains and activities within those chains. In addition, information on location, accident type, and different types of consequences (e.g. human health, environmental and economic impacts) is coded for to achieve a comprehensive global coverage of severe accidents. The database uses seven criteria to distinguish between severe and smaller accidents. Whenever one or more of the following consequences is met, an accident is considered to be severe:
  • at least 5 fatalities or
  • at least 10 injured or
  • at least 200 evacuees or
  • extensive ban on consumption of food or
  • releases of hydrocarbons exceeding 10,000 (metric) tons or
  • enforced cleanup of land and water over an area of at least 25 km2 or
  • economic loss of at least 5 million USD (2000)
Finally, comparative evaluations of energy technologies encompass a broad range of analytical methods, including:
  • aggregated consequence indicators
  • frequency-consequence (F–N) curves
  • advanced methods such as Bayesian networks, fat-tail distribution fitting, multivariate techniques and data mining using tree-based models,
  • coupling of ENSAD with Geographic Information Systems (GIS) to enable geo-statistical approaches and risk mapping,
  • economic loss and external cost estimation
  • evaluation of risk indicators within Multi-criteria Decision Analysis (MCDA) to provide support to decision processes

Historical Development of ENSAD Database

Besides these more technical and process-related developments, numerous extensions concerning the scope and analysis options have been accomplished in the course of specific research projects and related activities:
  • EU Project NewExt (2001-2003): External cost calculations of accident risks for non-nuclear energy chains. Estimation of uncertainties for results of standard methods; i.e. aggregated indicators and frequency consequence curves. Final NewExt Report (2004).
  • EU Project NEEDS (2004-2008): Trend extrapolation of specific risk indicators for a broad set of technologies to the year 2050. Application of the simplified Probabilistic Safety Assessment (PSA) for the nuclear chain to selected future designs. Coupling of ENSAD with Geographic Information Systems (GIS) and multivariate statistical analyses to assign accident risks to specific geographical areas, and to produce illustrative maps and contour plots showing spatial patterns. Calculation of specific risk indicators as input for Multi-criteria Decision Analysis (MCDA). Final report on quantification of risk indicators (2008).
  • EU Project SECURE (2008-2010): New methodological developments including the use of Generalized Pareto Distribution (GPD) and Bayesian Network analysis. Consideration of intentional human action, such as vandalism, sabotage and terrorist attacks within the broader context of critical infrastructure protection (CIP). Development and applications of a methodology for the assessment of the terrorist threat to major energy infrastructures. Final report on quantification of risk indicators (2011).

Selected Results

Import-adjusted fatality rates in the oil chain – reassigning responsibility
In this study, import-adjusted fatality rates were calculated for individual Organization for Economic Cooperation and Development (OECD) countries caused by accidents in the oil energy chain. A one-dimensional accounting method was used, based on trade data to determine the crude oil consumption fatality rates of the OECD countries annually between 1978 and 2008. This analysis results in meaningful changes to production based fatality calculations. In particular, OECD countries import the majority of their annual fatality rates from non-OECD countries. Figure 2 shows a selected result of this work for the United States (U.S.). Imports from non-OECD countries make up the majority of the adjusted fatality rate in most years. Interestingly, the U.S. has a historically significant fatality rate associated with domestic production, which has been declining with respect to domestic production over the analyzed 30-year period (1978-2008). However, these domestic accidents appear to be replaced by accidents from non-OECD countries. There is a decline in U.S. oil production over this period, complemented by an increase in imports from many other countries e non-OECD and OECD alike. The largest increases in imports come from Canada, Venezuela, Nigeria and Saudi-Arabia. One might expect the fatality rates of the United States to be higher relative to other countries due to the sheer amount of oil imports; however, when normalized by consumption, the values scale downward.

Risk Fig2.jpg

Figure 2: The annual fatality rate of the U.S. is graphed for the years between 1978 and 2008. The red bars indicate a source from the OECD, black bars indicate a source from the non-OECD, and blue bars indicate the accidents are the result of domestic accidents. Source: Lordan et al. (2015)

An aftermath analysis of the 2014 coal mine accident in Soma, Turkey
This analysis aimed to determine if events such as the Soma coal mine accident in Turkey in May 2014 (301 fatalities) should be expected, in a statistical sense, based on historical observations (1970-2014). The method proposed is based on Bayesian inference, which allows an explicit treatment of uncertainties. Figure 3 shows the result for the expectation analysis of a Soma type event for four different cases, i.e., OECD, OECD w/o Turkey, Turkey and USA. For OECD and USA the Soma accident is found to lie within the credibility intervals at the 99% quantile, which represent 1% of the accidents. This means that the Soma accident, in terms of fatalities, could be seen as an extreme case for the OECD group according to the historical observations. In the case of OECD w/o Turkey, an accident with the consequences of the Soma event becomes even more unlikely at the level of 0.1% of the accidents. Finally, the Soma accident is expected to be more likely in the Turkey dataset compared to the OECD and OECD w/o Turkey cases. In fact, the maximum consequence accident falls within the confidence interval at a level of 7% of the accidents.

Risk Fig3.jpg

Figure 3: Maximum observed consequence, e.g., Soma accident (triangle) compared to the confidence interval (5–95%) of the fatalities exceeded in X%, which is the probability level of accidents (1 – modeled quantile) where the maximum observed consequence falls in the modeled confidence interval, of the coal mine accidents (circle = mean) for the OECD, OECD w/o Turkey, Turkey and USA datasets. Source: Spada and Burgherr (2016)

Regionalized Risk Assessment of Accidental Oil Spills
The goal of this study is to assess regionalized oil spill indicators to compare and prioritize risk across worldwide regions, but also to identify potential hotspots. Figure 4 presents aggregated risk scores of maritime ship spills for 16 offshore regions. The regions with high spill risk include the Baltic Sea, EW South Atlantic, Caribbean Sea and NE Atlantic, whereas lowest risks were found in the Sea of Japan, Sulu and Red Sea. The other regions can be classified as low-medium (NW Pacific, S Pacific, GOM), medium (EW Indian Ocean, NE Pacific, Yellow Sea), and medium-high risk (E Med, Persian Gulf, W Med). The geometric interval classification was used to assign a region to a risk class. The advantages of this scheme are: (1) each class has about the same number of values, (2) balance between highlighting changes in the middle values and the extreme values, and (3) a visually appealing and cartographically comprehensive map is obtained.

Risk Fig4.jpg

Figure 4: Aggregated risk scores for maritime ship spills in 16 maritime regions, based on four indicators. Source: Burgherr et al. (2015)

Selected Publications