Systematic literature review on energy infrastructure resilience from a data analytics perspective
Harsh Anand1 and Mohamad Darayi2
1Data Science and Analytics, M.S. Program, 2Assistant Professor of Systems Engineering, School of Graduate Professional Studies, The Pennsylvania State University
1 Abstract
2 Introduction
3 Research Methodology
4 Taxonomy Overview
Critical infrastructures include any system involved in the production, transportation, communication, and operation of power to consumers for end-use. Energy infrastructures such as electric power, natural gas, and fuel networks are critical systems experiencing revolutionary shifts in all aspects. These systems are highly interconnected within themselves and other critical infrastructure to run the overall economy. With growing dependency, the energy infrastructure networks have become more complex, and their vulnerability under disruption is a prime concern for the smooth functioning of the economy. Each year the U.S. government has been spending millions of dollars to subside the risk of energy infrastructures under natural and human-made hazards such as hurricanes and terrorist attacks, which threats the overall operation of the infrastructure and the economy. It is of utmost importance for the central decision-makers to quantify and extensively evaluate the infrastructures.
To understand the challenges faced by the research community and the open problems in energy infrastructure resilience, we have studied the works of literature in two broad domains – systems modeling and analytics. Figure XXX gives an overview of systems modeling and analytical perspectives. The study is conducted for the readers to understand (1) analytical approaches to solve the critical challenges in energy infrastructure resilience and (2) different modeling techniques used to reach the objectives.
4.1.1 System Modeling
4.1.1.1 Interdependencies
4.1.1.1.1 Geographical
4.1.1.1.2 Physical
4.1.1.1.3 Cyber
4.1.1.1.4 Logical
4.1.1.2 Performance Measures
As the energy landscape shifts from a traditional grid composed of wires, generators, and growing demand into the grid of the future, all of those aspects of the traditional grid are being redesigned and incorporating new technologies and policies to shift with the world around it. The research is following these shifts, most notably in two distinct ways: sustainability and resilience.
4.1.1.2.1 Sustainability
Sustainability, generally, encompasses a longevity of the industry and its ability to adapt to changes. In the energy, it can be looked at through four lenses: environmental, technological, economic, and social. Each of these aspects have become more notable in literature since the implementation of the modern grid decades ago is now aging and society has different focuses for its future.
Environmental:
Renewable resources emerged quickly over the past couple of decades. Climate change has not only put pressure on the planet, but it has also caused consumers to be more proactive in their energy consumption.
Literature aiming to evaluate the environmental sustainability of a SoS includes studies to minimize household carbon emissions, such as the objective of Agusdinata and Dittmar (2009), who consider the impacts of both the supply side and demand side in their household emissions SoS. However, other researchers evaluated environmental sustainability much more broadly. Hadian & Madani (2014) created a framework to evaluate the environmental impact of, carbon emissions, land use, and water use of electric generation fuel types as they show in Figure 2. They also considered the cost of the electric generation to better evaluate the tradeoffs of each resource type. This research was then replicated by Ristic et al (including Madani) in 2019 to evaluate fuel alternatives in the European Union as they seek to decarbonize.
Economic:
Third, economic sustainability requires the energy industry to be efficient in the cost to consumers, producers, and government bodies. This lens can often be a large hurdle in the process of retrofitting or retiring existing resources, implementing new resources, and continuing to provide power at the least cost, both now and in the future.
Mostly all literature which had a strong focus on economic sustainability also had a strong objective for one other classification of sustainability. Most notably, technological sustainability and economic sustainability were paired together including the works of Allman and Daoutidis (2016), Ristic et al (2019), and Arasteh et al. (2019). However, some research focused more on the sole economic aspects such as Mittal et al (2015), whose framework focused on analyzing market retail structures and Moloney, Fitzgibbon, and McKeogh (2017) who’s framework focused on sustaining infrastructure. Their work also classifies in the resilience objective.
Social:
Finally, social sustainability is the least intuitive out of the four aspects of sustainability. It focuses on the social benefit of the energy system. In other words, its primary goal is the make the collective society and the people within it better off. Though that is difficult to quantify, the measures of it often contain the other three aspects of sustainability depending on how someone defines the social benefit, including least cost options, increased availability of resources, or policy-making, amongst others.
The social aspect of sustainability was the least common trend in discrete objectives. However, each of the other three aspects can inherently provide social benefit as well. Xiao, Hipel, and Fang (2019)’s objectives focused on social sustainability the most in their analysis of the water-food-energy nexus as society needs access those vital resources to survive. Others, including Hadian & Madani (2014) and Ristic et al (2019), acknowledged the societal benefit their research could provide in less direct ways.
4.1.1.2.2 Resilience
4.1.1.3 Mechanism
4.1.1.3.1 Deterministic
4.1.1.3.2 Stochastic
4.1.2 Analytics
Analysis in systems thinking, like many other fields, refers to how the relevant data is interpreted, utilized, evaluated, and predicted in the context of the problem and at any point of the problem. For instance, analytics can be used to determine the inputs to the system-of-systems framework, it can be used to evaluate the outputs of the framework, or it can be used to predict future success, amongst many other applications. Analytics, broadly, are categorized as descriptive, prescriptive, and predictive.
4.1.2.1 Descriptive
Descriptive, as the name implies, describes the data as it exists or previously existed; its types include mean, variance, correlation, and others. Descriptive analytics can provide valuable insights into the past. This can help people make educated decisions and potentially utilize other analytical techniques. Descriptive analytics is utilized in in most data-based studies to quality check data, verify reasonableness of models, or evaluate results; the SoS literature is no exception.
Almost all literature reviewed for this research contained some aspect of descriptive analytics. This is especially present when the researcher uses predictive techniques as well. The predicting methods must use historical data to learn and the researchers must first describe the data they intend to use to ensure the features are valid and relevant to the predictions. For example, Agusdinata and Dittmar (2009) determined plausible ranges, using descriptive statistics, to simulate observations for their classification and regression tree prediction module.
4.1.2.2 Predictive
Predictive analytics uses existing data, mathematical formulations and algorithms to predict the future behavior of data. Predictive analytics has more complexity than descriptive and lends itself to much more in-depth fields of study, such as machine learning. Like many other industries, predictive analytics is becoming more accessible with the large amounts of data as well as readily available tools to analyze the data using robust methods and advanced techniques.
Although the study of predictive analytics has emerged rapidly, its presence is not common in the study of SoS frameworks due to the complexities of both predictive techniques and the SoS itself. However, some researchers utilized the techniques in their studies, though often it was not the main focus of the research. Most notably, forecasting techniques were used to predict external behaviors (Agusdinata and Dittmar, 2009; Mittal et al., 2015; Sianaki and Masoum, 2014); Sianaki and Masoum (2014) have included an entire predictor system which predicts five different elements of the analysis, including demand quantities and timing, energy prices, and the availability of renewable resources.
4.1.2.3 Prescriptive
Prescriptive analytics can utilize the information provided from descriptive and predictive analytics to inform decisions or provide recommendations. Methods of prescriptive analytics include optimizations such as linear programming or stochastic optimization.
Optimizations are very often used to approximate optimal power flows, especially in regards to the energy markets. Therefore, many of the reviewed literature contains some element of prescriptive analytics. Specifically, Zhao et al. (2018) developed a bi-level optimization model which minimizes cost of the SoS subject to various uncertainties. Their study focused on microgrids and their interaction with each other and the larger distribution grid. Similarly, Thacker, Pant and Hall (2017) used a capacity constrained location-allocation optimization algorithm in their framework.
5 Analysis and Discussion
6 Conclusion
6.1 Contribution of this study
6.2 Limitations
6.3 Research Scope
