DIGITAL EDITION

 

The current issue of ORMS Today is now available to INFORMS members in "digital magazine" format. Log in to the Digital Editions Archive and enjoy a new online reading experience.

Software Surveys

  STATISTICAL ANALYSIS SOFTWARE SURVEY
Biennial survey of statistical analysis software...
more »

O.R.: Catalyst for Grand Challenges

Opportunities in sustainability

By Suvrajeet Sen, Cynthia Barnhart, John R. Birge, Warren B. Powell and Christine A. Shoemaker

sustainability operations research

Image © Rafomundo | 123rf.com

Editor’s note:
This is the second in a series of articles based on a report to the National Science Foundation titled, “O.R. as a Catalyst for Engineering Grand Challenges.”
The Earth is a planet of finite resources, and its growing population currently consumes them at a rate that cannot be sustained. Utilizing resources (like fusion, wind and solar power), preserving the integrity of our environment and providing access to potable water are the first few steps to securing an environmentally sound and energy-efficient future for all of mankind. Analyzing these challenges includes using data and models to choose among alternative strategic decisions, forecasting the effect of decisions on the future, and quantifying the uncertainty associated with this analysis. While O.R. methods certainly support many individual features (e.g., nonlinear, dynamic, stochastic or discrete), combinations of these features are often necessary in many of the research questions that arise under the “sustainability” banner. Combining these features will require significant extensions of the O.R. methodology available today.

What makes O.R. vital for the future of sustainability?
O.R. methodology such as applied probability, decision analysis and optimization have been central to oil and gas exploration, blending and transport since the inception of our discipline. The same has been true of the role of O.R. for operational problems arising in planning electric power generation and transmission, natural gas transport and others. Similarly, the use of O.R. in water resources, and environmental planning are well documented in the literature. However, the growing evidence of global warming is not only shifting the emphasis on green technologies (e.g., higher fuel efficiency standards), but also in our fundamental understanding of processes to curb the increasing volatility of global climate. For instance, the recent National Oceanic and Atmospheric Administration (NOAA) report, “The State of the Climate in 2012,” noted that the year 2012 was one of the 10 hottest years on record (global average), and moreover, a vast majority of the top 10 hottest years have been recorded since the 1990s [1].

In connection with this report, NOAA’s acting administrator, Kathryn Sullivan, mentions that “many of the planning models for infrastructure rely on the future being statistically a lot like the past, and certainly the data should lead one to question if that will be so.” In other words, future planning models ought to recognize on-going changes due to human activity. This calls for a framework that seamlessly integrates data and decisions, and ultimately sheds light on one or more hypotheses through a validation exercise. This interplay leads directly into the emerging world of integrated analytics, which is envisioned in Figure 1. Because of its core competencies in all facets of integrated analytics, INFORMS is in a unique position to lead such thrusts to enhance sustainability.

This integrative orientation is particularly relevant for meeting sustainability challenges. Because environmental phenomena (e.g., temperature, humidity, pressure systems, tides, wind, etc.) exhibit significant spatiotemporal correlations, stochastic intermittency and nonlinearity, our attempts to harness natural resources must be undertaken with great care. Fortunately, the growth of inexpensive sensors, in conjunction with communications networks, provides a massive data infrastructure that can be leveraged at various levels of granularity, ranging from streaming regional data to worldwide data at coarser granularity.

Framework for integrated analytics.

Figure 1: Framework for integrated analytics.

Energy

In order to become less reliant on fossil fuels, significant attention is being paid to increasing the percentage of renewable energy from a variety of sources. This transformation will require innovations ranging from devices and processes for high efficiency photonic crystals to improved integration of renewable resources into the power system. O.R. methodology is beginning to appear in a variety of settings, ranging from new devices (e.g., photonic crystals using band gap optimization) and manufacturing processes, to novel ways of operating large systems with significant renewable penetration.

Integrating storage with wind and solar
Most states in the U.S. have mandated that at least 20 percent of power supplied by a utility must be generated by renewable energy by 2020. In some cases, such as California, the law will require utilities to supply at least 50 percent of power generation from renewable sources by 2030. In order to achieve these mandates, it will be important to manage the variability using a balanced portfolio that combines careful use of dispatchable fossil generation, potential use of biofuels, demand response and storage.

Renewable energy offers the best potential for zero CO2 emissions, but considerable research is being devoted to the challenge of solving the problem of dispatchability, which refers to our inability to control wind and solar production. Considerable attention has also been devoted to the idea of using storage to smooth the variations, but the tools of operations research have highlighted some problems with this simple solution.

Storage behind the meter (that is, in a residence, commercial building or microgrid) is very expensive on the margin, since the last unit of investment is only used a small amount of time (grid operators are well familiar with this problem and have devised numerous strategies to help mitigate peak usage). Grid level storage (in front of the meter) is even more problematic; even if it were free, the conversion losses make it more economical to ramp dispatchable sources such as gas generators than to incur the 15 percent to 20 percent conversion losses to charge and discharge batteries. It is the tools of operations research that have identified these issues and are helping to inform the debate over how to handle the variability of renewables by designing and controlling portfolios of energy generation, demand response and storage, tuned for each region of the country.

A major problem faced by battery operators is monetizing the ability of storage to smooth the variations from wind and solar over larger time scales. Surprisingly, finding optimal policies for managing energy storage resources is beyond the state of the art of modern algorithms for all but the most trivial systems. Real battery systems have to deal with dynamic energy generation from the wind and sun, stochastic prices, time varying loads and complex battery chemistry that affects the efficiency of energy conversion and battery lifetimes. Particularly challenging is optimizing generation and storage over multiple time scales, from responding to frequency regulation signals every few seconds to the hourly and daily variations that characterize diurnal and weekly cycles.

We do not yet have the tools to obtain optimal policies for realistic battery systems. Standard practice in engineering is to live with various heuristics, but these can perform well below optimality. Proper battery management, especially for affordable technologies such as lead-acid batteries, can dramatically extend the lifetime of the device. The best policies can outperform “good” policies by 100 percent. Furthermore, it is likely that these algorithmic advances would benefit both existing and future storage technologies.

The next generation of optimal storage control algorithms needs to handle a number of problem characteristics that are beyond the reach of existing tools: a) realistic battery chemistry, b) batteries cognizant of the “state of the world” (temperature, loads, prices, etc.), c) batteries coupled as subsystems within a larger portfolio of storage devices, and d) batteries timing their charge/discharge policies to coordinate with the market.

Environment

In the interest of brevity, we will focus on two issues: geological carbon sequestration and managing water availability and quality.

Geological carbon sequestration
The rate of increase in carbon dioxide emission may be the single biggest culprit among all greenhouse gases. It is predicted that if these levels continue to rise, the consequences could be severe: rising sea levels, disruptions in agriculture and other natural disasters (e.g., hurricanes, tornadoes) may begin striking with greater force and frequency. Some possibilities for geological carbon sequestration (GCS) include storing the gas in deep saltwater aquifers or depleted oil reservoirs. However, there are potential problems, both economic and environmental, with GCS. The environmental risks are that injected CO2 could increase subsurface pressure to the point that CO2 or saltwater would move upward through fractures or defective well casings. Such movement could result in pollution (including increased metal concentrations) of freshwater aquifers used for human drinking water or in the escape of CO2 into the atmosphere, which, at worst, could cause deaths and, at best, represent a waste of the large amount of money spent to sequester the CO2.

Solutions for long-term sustainability will no doubt include solar- and wind-powered energy sources. Image © Vaclav Volrab | www.123rf.com

Solutions for long-term sustainability will no doubt include solar- and wind-powered energy sources. Image © Vaclav Volrab | www.123rf.com

O.R. methods can assist in promoting safe geological sequestration of carbon by helping to design monitoring systems and using optimization and statistical methods to estimate the current and future spatial distribution of CO2 and pressure in the injection region. Note that these injection regions are expected to be at least 1,000 meters below the surface. Because the injection site is so deep, the monitoring wells also need to be very deep and are therefore extremely expensive. As a result, a practical monitoring system might consist of only a few strategically located monitoring sites, some of which may be in regions outside the CO2 plume to monitor pressure without increasing the risk of CO2 escape. Since the CO2 will take the shortest path upwards, knowing the location of the CO2 plume is very important, and sensor locations should be chosen to identify leaks reliably.

Recent work has demonstrated that a combination of a three-dimensional simulation model of CO2 can be used with nonlinear optimization methods and uncertainty quantification method to obtain credible estimates of current spatial distribution of the CO2 and even produce good predictions of its future migration. This is a very difficult problem because the simulation models for a reasonably sized CO2 geological storage site (and other multiphase subsurface fluid problems) are highly nonlinear and can take many hours – or days – for one simulation.

Managing water availability and quality
Approximately one in six people living today do not have adequate access to water, and more than double that number lack basic sanitation. Indeed, lack of clean water is responsible for more deaths worldwide than war. Providing access to clean water is an engineering challenge of the highest priority. The goals are to protect water from pollution so that it can be used to meet human and ecological needs and to ensure that the quantity of water is adequate. Insufficient quantity of unpolluted water leads to disease, agriculturally and industrially devastating drought, and loss of valuable renewable hydropower. Such questions involving social welfare, especially the interface with technology, have provided fertile grounds for O.R. applications in the past, and we expect their impact to continue in the context of this worldwide challenge.

Most environmental studies of pollution and ecological health are based on either mathematical or computer simulation models known as “flow and transport” models. One instantiates such models by fitting parameters to data (e.g., soil characteristics, weather, pollutants), and using the output of the model to predict the dynamic and spatial distribution of water and water-borne pollutants. These predictive models can then be used in a variety of ways, including flow and transport models, as well as trade-offs between alternative remediation policies.

While such methodology is not foreign to the water resources community, the large scale and computationally demanding nature of these models make them worthy of collaboration. To give a sense of some of the challenges, consider a case of designing a network for water purification and delivery. Given that such processes can cost more than a billion dollars to set up, they must be placed strategically, and the network itself must also be designed to be cost effective. The question of delivering clean water can be posed as a decision (optimization) model to make some discrete choices (e.g., location and network design) so that adequate quantities of clean water (modeled by continuous variables) can be delivered cost effectively. Such models, leading to nonlinear mixed integer programs, can be notoriously difficult, and O.R. expertise is needed to solve what is currently an intractable problem. Ultimately such advances will reduce costs while increasing the quantity of clean water available for use.

A major challenge and recurring theme in this setting is the need to solve, design and control problems in the presence of different forms of uncertainty. Design of water purification technologies, the location of purification facilities, the allocation of sensors and metering devices and the management of water reservoirs all represent different forms of stochastic optimization problems that have to be solved in the presence of different forms of uncertainty.

Social and Economic Sciences for Sustainability

Much of the emphasis in sustainability efforts relates to social interactions for which O.R. models can provide guidance. Basic economic reasoning suggests, for example, that solar energy prices will decline relative to the prices of other energy resources if new solar technologies allow for greater efficiency than currently exists or if the capital cost of producing energy with current solar technologies is reduced. Either of these outcomes requires investment in new conversion technologies or in the devices and manufacturing processes for current technologies. Given these alternatives, solar energy should become more economical if investment increases in developing new technologies of either form.

O.R. methodology can help in assessing the value of policies to encourage such investment. In particular, O.R. models of interactions between the economy and the energy sector can be used to assess the effectiveness of policy alternatives by evaluating overall social costs measured, for example, in changes in GDP growth, effects on specific industries or trading imbalances. Such O.R. models have a long history of informing government policy. Similarly, equilibrium models for energy production and consumption have a long history from the late 1970s and have evolved into more recent policy questions such as the economic impact of greenhouse gas emissions.

The second sustainability challenge of improvement in the power grid can be most effective if individual consumers can learn to modify behavior to shift consumption in response to current conditions. O.R. models can be used to evaluate responses to policies such as dynamic pricing.

Finally, O.R. models can help provide access to clean water by aligning policies with long-term goals. The allocation of water rights and policies to manage those rights again raises the issue of balancing the preferences of individuals with those of society as a whole. O.R. models can analyze these decisions through representations of preferences and the mechanism of coordination.

Reference

  1. For more recent climate data, see https://www.climate.gov/maps-data