Balancing cost, water, emissions, and reliability in power systems operations

Traditionally, large-scale thermoelectric power generation has been operated to reduce system operational costs. To expedite the mitigation of the harmful effects of climate change, many have proposed additional incentives for system operation (i.e. policies) that incorporate greenhouse gas emissions. However, such policies rarely consider unforeseen impacts on the volumes of water required for cooling thermoelectric plants as well as the potential effects on electricity production from water/climate-related stressors. We first create a case study representative of the thermoelectric-dominated water/energy systems in the Midwestern United States. Through this case study, our analysis investigates the tradeoffs of cost, water, emissions, and reliability in thermoelectric-dominated water/energy systems via policy analysis. Furthermore, we show how such policies respond differently to historic operational, climatological, and hydrological stressors. Specifically, we find that policies that focus on a single criterion can leave power systems vulnerable to reliability issues, operational cost increases, ecological impacts on riverine systems, and increased emissions. Therefore, consideration of many criteria (cost, water, emissions, and reliability) is necessary for creating an effective water-energy-emissions policy.


Introduction
Energy systems are at the intersection of a functioning economy, a clean atmosphere, and healthy waterways.The current economic value of US power plants is approximately $1 trillion (US dollars) (Rhodes 2017).Globally, 25% of emissions released into the atmosphere come from electric power production (Hockstad and Hanel 2018).Even as many power systems transition to renewable generation sources, it is currently estimated that 89% of energy sector greenhouse gas emissions are a result of fossil fuel combustion and industrial processes (IEA 2021) with a similar percentage of total power coming from such sources.In the United States, an estimated 48% of fresh surface water withdrawals (water that is diverted from a source) can be attributed to thermoelectric power plant electricity generation (Dieter et al 2018).Although the water consumed (water that is evaporated) is far less, proposed technological solutions to reduce emissions, such as carbon capture systems, have been shown to increase water consumption (Zhai and Rubin 2010).
The impacts of climatological grid stressors expose how power plants are at the heart of this intersection-and how this intersectionality often impacts the reliability of energy systems.Droughts and heat waves increase the temperature of cooling water and electricity demands-thus impacting energy reliability, increasing water withdrawal and consumption, creating negative impacts on riverine ecology from elevated-temperature discharge water, increasing operational costs, and relying on higher-emitting fuel sources (Poumadére et al 2005, Galbraith 2011, Reuters Staff 2011, Eaton 2012, Freedman 2012, Harto et al 2012, Luo 2017, Lubega and Stillwell 2018b).Impacts on grid reliability during heat waves are particularly concerning as such coupled stress events increase mortality and morbidity risk (Stone et al 2023).In many regions, climate change increases the severity and frequency of these coupled stress events.
During normal operations, power systems operations have the objective of minimizing operational costs while satisfying operational constraints on grid reliability, thermoelectric discharge temperature regulations, and emissions regulations.However, if grid stressors are extreme enough, these constraints may be violated.For example, system operators often choose between enforcing existing policies that limit the temperature of discharged water or granting thermal variances that allow discharge beyond legal limits (Poumadére et al 2005, Harto et al 2012, Lubega and Stillwell 2018a).Thermal variances are almost always granted (Micha 2014), often due to concerns of impacting grid reliability.However, even grid reliability is not always achieved during particularly stressful climatological events.Therefore, cost, reliability, emissions reductions, and avoidance of thermal discharge violations are, in reality, ideals that compete with one another.Thus, to create operational policies that effectively balance impacts to grid reliability, thermoelectric discharge temperature violations, and emissions under extreme grid stressors, it is more effective to model these quantities as competing objectives rather than hard constraints.
Previous studies have demonstrated that modeling power system emissions, water use, and reliability as extra constraints or objectives modify operations.However, previous frameworks have studied such quantities in a two-at-a-time fashion-for example, cost-emissions (Abido 2003), reliability-water (van Vliet et al 2016), cost-reliability (Billinton and Allan 1984), and cost-water over a variety of planning timescales.Water use has been effectively incorporated as an objective and constraint in long-term power system planning problems (Jornada and Leon 2016, Lubega and Stillwell 2018a, Liu et al 2019, Mu et al 2020) as well as the daily and sub-daily operations of power systems (Sanders et al 2014, Fooladivanda andTaylor 2015).However, there is a need to holistically create new cost-water-emissions-reliabilityinformed operations.These measures promote system coordination rather than manual intervention by system operators, as well as offer a rapid way for system operators to respond to current drought events, especially if long-term solutions are infeasible due to a lack of time to implement (Pacsi et al 2013).
This work contributes a tradeoff analysis of cost, water, carbon dioxide emissions, and reliability of energy systems, via proposing policies.A policy in this paper consists of three decisionsthe per unit costs assigned to water consumption, water withdrawal, and carbon dioxide emissions.Such costs allow water use and emissions to be directly incorporated into existing power systems operations and are an extension of previously proposed water/energy policy frameworks (Kravits et al 2022b).We couple existing models of thermoelectric power plant water use (Rutberg et al 2011, van Vliet et al 2016), emissions rates, and power systems reliability in the real-world operational problem of optimal power flow (OPF), which optimizes generator power output based on a minimization of operational costs.Additionally, we apply our methodology to a representative thermoelectric-dominated water/energy system in the Midwestern United States.This area was strategically chosen due to its historic issues with thermoelectric discharge temperature violations leading to electricity unreliability (Freedman 2012, Lydersen 2016).Through this case study, we use a multi-objective evolutionary algorithm (MOEA) to generate policies that efficiently compromise among power system cost, water, emissions, and reliability.A tradeoff analysis of this set of policies allows us general insights into water/energy system interplay as well as allows us to select a single 'wateremissions' policy that effectively balances all quantities.This balanced policy shows improved performance relative to policies solely created using water or emissions improvements, thus demonstrating how effective energy policy would benefit from a multiperspective approach.

Policy decisions
A 'policy' in this study is comprised of three 'policy decisions' .They are the operational penalties assigned to emissions, water consumption, and water withdrawal.These operational penalties act as objective weights that allow us to convert the quantity of emissions, water consumption, and water withdrawal into the monetary cost of emissions, water consumption, and water withdrawal.These monetary costs are then introduced as additional objective terms when computing the monetary cost of operating the system.For example, a water withdrawal penalty with the units of dollars per gallon of water allows us to define a water withdrawal cost term to be directly added to other operational costs.Such a policy formulation is an extension of that proposed in Kravits et al (2022b).
Each unique policy expresses a relative preference for reducing water withdrawal, water consumption, and emissions.For example, if a policy specifies a penalty on water withdrawal too low, there will be no significant reduction in system water withdrawal.However, if the value is too high, then the system drastically reduces water withdrawal at the expense of significant increases in operational costs, emissions, and water consumption.Therefore, it is crucial to make an informed policy decision that holistically considers all aspects of energy system performance.

Water/energy/emissions cosimulation
Through creating a week-long hourly-resolution water/energy/emissions cosimulation model, we capture the system-level and generator-level performance impacts of policy decisions.The water component of the cosimulation captures the water withdrawal and consumption of the generators.The emissions component captures the carbon emissions associated with energy production.Finally, the energy component captures the underlying generator operations of the electrical power system.

Water use simulation
To simulate the water withdrawal and consumption (i.e. the ratio of water used to energy produced; gal MWh −1 ) of once-through-cooled plants, we implement models from Rutberg et al (2011).Once-through systems provide cooling by directly warming colder withdrawn water by running it past a heat exchanger.Thus, the original model proposed in Rutberg et al (2011) uses the energy dissipated by the power plants to determine the discharge rate and temperature difference between inlet and outlet temperatures.Additionally, there are empirical correction factors to account for the complexities and inefficiencies of real-world generators.However, there is an upper regulatory limit on the temperature of water that can be legally discharged.Thus, we extend their models such that each plant stays within its legal operating temperature bounds unless it hits its maximum discharge rate based on the assumption laid out in Langford (1990), Madden et al (2013).After hitting its maximum discharge rate, temperatures can exceed legal limits as has been historically done in the case of thermal variances (Lydersen 2016).The importance of this discretized behavior for creating optimal policies is discussed later in section 3.3.For additional details of how these models were combined, see SI note 1.1.1.
Recirculating-cooled plants provide cooling by using cooling towers to facilitate convective heat transfer to ambient air.Therefore, the withdrawal and consumption of recirculating systems are most sensitive to the ambient air temperature (Rutberg et al 2011, Diehl et al 2013).Thus, we use the models proposed in Rutberg et al (2011) to model water-use rates of recirculating plants.See SI note 1.1.3for additional details of the simulation's implementation.
Another consideration when simulating a plant's water use is the reduction in a plant's energy generation capacity.These limited capacities are particularly relevant during extreme heat and drought events where, even if a plant's cooling system is operating at its highest withdrawal or consumption rate, the plant must operate at a lower output capacity to ensure safe operations.To capture the impacts of climatology on generator capacity, we extended the models from van Vliet et al (2016) (SI note 1.1.3).Specifically, we use the simulated water withdrawal/consumption rates previously described as inputs to the generator capacity reduction models.Additionally, this model considers key hydrological details such as the streamflow of the rivers used as a source of cooling water.
The result of the combination of these models is a high-fidelity plant water-use simulation.Given system air temperature, streamflow, stream temperature, and various hydrologic/operational correction factors, our simulation allows us insights into the quantities of water used, the temperature of discharged water, and possible diminished generator output capacity.

Emissions simulation
This simulation focuses on carbon emissions due to their globally increasing trends in the energy sector (Friedlingstein et al 2022).Fossil fuel sources have corresponding carbon emission factors (e.g.lbs of CO 2 per MWh).Estimates of these values are typically based on historical information and are generally well-accepted (EPA 2021).Such factors are largely determined based on the type of fuel source.Therefore, our cosimulation captures total system emissions as the product of a generator's energy output and its emission factor.

Power system simulation
The core of the simulation model is a water-and emissions-informed DC OPF model.The traditional DC OPF operational problem is an optimization problem solved by system operators within electricity markets.The decisions of this problem are the dispatched power levels at every generator in a system.Dispatched levels are constrained such that both the power flowing on the lines and the generator capacities operate within safe limits.The objective of the traditional DC OPF problem is the minimization of operational costs (e.g.fuel costs).
To create our water-and emissions-informed DC OPF, we introduce additional objective terms for the costs associated with total water withdrawal, water consumption, and emissions.These terms are the product of the simulated water withdrawal, water consumption, and emissions (from sections 2.2.1 and 2.2.2) and their respective operational penalties (from section 2.1).See SI note 1.2 for the explicit formulation of our water-and emissions-informed DC OPF.
The reliability of this system is also modeled as an additional penalty term in the objective of our water-and emissions-informed DC OPF.Much like the operational penalties, the value-of-lost-load term specifies a monetary price associated with not serving electricity demand.These monetary prices are typically expensive to avoid interruption in electricity supply (Anderson et al 2019).Thus, the system will only choose not to serve demand if either all generators have reached maximum capacity (so the system has no choice but not to serve demand) or, in the case of our water-and emissions-informed DC OPF, the operational penalties on water use or emissions are priced so expensive that it is cheaper not to serve demand.

Multi-objective optimization
Six objectives each capture a different aspect of a decision-maker's value system.The cost objective captures the economic cost of operating the system and is largely determined by fuel costs and generator power output.The water withdrawal and water consumption objectives are the product of the simulated water use rates and dispatched energy output summed over the simulation period (see section 2.2.1).Similarly, total carbon emissions are the product of generator output and emission factors (see section 2.2.2).The reliability of the system is captured by modeling the energy not served in the system over a simulation time horizon (see section 2.2.3).Both the quantity of water as well as the level of temperature violation is captured in the discharge violation objective.See SI note 1.3 for each objective's explicit formulation.
Policy decisions that only consider a single objective have the potential to adversely impact other objectives (as discussed in section 2.1).We use multi-objective optimization to create a nondominated set of policies-a set where each policy uniquely balances our objectives of interest.Creating effective energy policies that balance the competing interests of decision-makers is imperative as it ultimately impacts the production scheduling of energy systems.

Tradeoff analysis
Once we have our nondominated set of policies, we conduct a tradeoff analysis to first understand the general interplay between our objectives.This analysis allows us general insights into the systemic behavior of water/energy systems.For example, we understand which objectives compete with one another and which are harmonious.
In the final phase of our framework, we select a single representative 'water-emissions' policy that seeks an equal compromise with respect to several objectives (cost, water withdrawal, water consumption, emissions) from the nondominated set.We simulate its performance under different system stressorseach selected to capture a different historic stressor of thermoelectric-dominated energy systems.Through this simulation, we understand the sensitivity of the performance of various policy decisions as well as how an optimal energy system policy can mitigate the impacts of various stressors.

Case study
The following section describes the specific assumptions made when applying our methods to a case study via a computational experiment.

Region
We apply our framework to the Illinois 200-bus synthetic grid currently hosted in the Electric Grid Test Case Repository (Birchfield et al 2017), a region that has historically been prone to thermoelectric temperature discharge violations.This 180 line transmission system has a 2800 MW capacity with half of its capacity coming from coal and the remaining capacity being equally split between natural gas, wind, and nuclear.Cooling systems have been assigned to these generators following the methods proposed in Kravits et al (2022b).Generator-specific parameters are sourced from Myhre (2002), James et al (2019), and the Energy Information Agency Form 923.We set a maximum regulatory discharge temperature for once-through generators to be 90

Data sources
For the ambient air data, the National Solar Radiation Database location near Springfield, Illinois from 2015 to 2020 (Sengupta et al 2018) is used.For this system, the streamflow temperature data do not have consistent temporal coverage.To supplement these limited data, empirical air temperatures are fitted to the stream temperature model proposed in Mohseni et al (1998), which has been applied to similar thermoelectric modeling problems (Koch andVögele 2009, van Vliet et al 2016).
The default bus loadings are supplied within the synthetic grid.Publicly available MISO data reflect the general system-wide trends in loads (e.g.impacts of seasonality).Bus-level hourly variability profiles are introduced following the methods from Li et al (2021).Wind capacity factors are applied based on historic wind speeds.

Computational experiment
We apply the multi-objective policy search and tradeoff analysis previously outlined in section 2. Although the six objective functions are convex, our simulation is highly nonconvex due to the underlying equations and discretized states as previously discussed.To optimize policies with respect to many competing objectives, given a highly nonconvex simulation, we choose to use an MOEA given the ability of modern MOEAs to pareto-approximate solutions to high-dimensional nonconvex problems (Hadka et al 2012, Hadka andReed 2013).We analyzed hypervolume progress throughout our search to ensure our solutions were both diverse and wellperforming.These methods are well-accepted when using MOEAs and have been outlined in Zatarain Salazar et al (2016) (For further details see SI note 3.3).The final product of our multi-objective policy search is a 'water-emissions' policy that balances the performance of each objective for an average week.
For comparison with our 'water-emissions' policy, we also generate four unique policies using single-objective optimization (table 1).These comparison policies have been chosen such that each focuses on a single aspect of water-energy-emissions system coupling.For example, the value of the water withdrawal penalty for the 'high water withdrawal penalty' policy is determined by increasing the penalty until no additional reductions to water withdrawal were observed.An analogous process is followed for the 'high water consumption penalty' and the 'high emissions penalty' policies.The final 'status quo' policy reflects current economic-only operations.These four policies ignore the coupling of water-energy-emissions systems as they each focus solely on a single objective at a time.Thus, they will be used as a comparison to the holistic 'water-emissions' policy.
We test our policies under several representative historical stress scenarios to observe the sensitivity of the performance of each scenario.The average week scenario is simulated as a baseline reference for how the system behaves when all the generators are available under average loading (the historic seven day period whose average load was closest to the long-running average load).The extreme load/climate scenario captures the week that had the highest seven day average air and water temperature.Such extreme weather produced heavy loads.The nuclear outage scenario simulates a nuclear outage either due to operational constraints or as a result of policy decisions.The line outage scenario simulates a planned or unplanned line outage under average loading.The avoid temperature violation scenario assumes that plant operators reduce capacity to avoid temperature violations under extreme load/climate.Thus, all scenarios besides the extreme load/climate and avoid temperature violation have the same load and climatology as the average week scenario.

The interplay of cost, water withdrawal, water consumption, and emissions
By comparing the relative performance of each policy for an average week, we gain insights into the general interplay of cost, water withdrawal, water consumption, and emissions.The performance of each policy is depicted in figure 1 such that policies intersecting axes lower have better performance (i.e. down is the preferred direction).An analogous plot for power output grouped by fuel mix and cooling type can be found in SI figure 10.Each policy increases operational costs and causes drastic changes to the fuel mix compared to 'status quo' operations but offers benefits such as reduced water consumption.Some policies offer focused benefits with respect to only a single objective while sacrificing performance in other objectives.For example, the 'high water consumption penalty' policy reduces status quo water consumption by 80% but increases status quo emissions by 120%.Policies that consider emissions ('high emissions penalty' and 'water-emissions') highlight that it is possible to reduce a system's water withdrawal, water consumption, and emissions at a marginal (5%) increase in cost.
This analysis reveals a nuanced tradeoff between water use, emissions, and cost.The general tradeoff between water consumption and water withdrawal has been established (McCall and Macknick 2016, Kravits et al 2022a), but the impacts of this tradeoff on operational costs and emissions have not been demonstrated.Generator cooling technologies that are water-consumption-efficient tend to be waterwithdrawal-inefficient and vice versa as previously discussed in section 2.2.1.The two policies that attempt only water reductions ('high water withdrawal penalty' and 'high water consumption penalty') have poor cost and emissions performance because they shift the fuel mix of the system toward costly water-efficient and emission-inefficient generators (e.g.natural gas combustion turbines seen in 1. Tradeoffs in policy performance for an average week.Status quo operations de-emphasize water consumption, withdrawal, and emissions.However, through multi-objective policy optimization, we find holistic policies that effectively compromise among all objectives.SI figure 10).However, the way that these policies achieve this performance is different; in our case study, each policies' relative emphasis determines the extent to which the system relies on expensive natural gas generators with recirculating cooling systems or coal generators with once-through cooling systems.Therefore, even though these two policies have similar costs, one produces the minimum withdrawal while the other produces the maximum withdrawal of the policies tested.However, this minimum/maximum behavior was not observed for system water consumption.Instead, the policy that minimized withdrawal had the fourth highest consumption.For our system, this nuanced behavior was due to the inclusion of a nuclear generator whose cooling system was inefficient with respect to both water withdrawal and consumption.This behavior led to 'status quo' operations exhibiting the highest consumption.Unfortunately, the severity of this tradeoff would be further exacerbated if generators install emission-capturing devices without considering how these devices further increase water consumption (Zhai and Rubin 2010).Clearly, how policies impact the interplay of cost, water withdrawal, water consumption, and emissions in energy systems, is both nuanced and critical for the future of our energy systems.

Comparing climatological and operational stressors
Modern energy grids, and the policies associated with them, must respond to various climatological and operational stressors (Pfenninger et al 2014, Miara et al 2017).The policy performance (bars) over each scenario (column) for every objective (row) is depicted in figure 2. We compare our policy's performance for several scenarios with varying degrees of climatological and operational stressors (table 2) to the average week performance previously discussed in section 4.1 (see SI figure 9).
Operational stressors not commonly associated with water/energy interactions, like a nuclear outage, can have more pronounced impacts on systems than the traditionally considered climatological stressors.The nuclear outage scenario leads to increases in operational cost, water withdrawal, and emissions compared to the extreme load/climate scenario under current operations.Physically, this result is because the system has lost an operationally inexpensive and lowemitting fuel source.This finding suggests that, for our case study, a nuclear outage is a larger stressor than historical climatological extremes.
Other uncommon water/energy operational stressors, like a line outage, can also impact policy performance.Physically, the transmission network topology constrains generator dispatch, although the critical line outage scenario led to slight (roughly 10%) increases in operational cost, water withdrawal, and water consumption-while emissions saw a greater impact.Particularly, 'status quo' emissions saw a 20% increase.However, the largest impact was on system reliability as a result of the 'high water withdrawal penalty' policy.Such behavior further highlights the importance of topological consideration in creating a holistic energy policy.
Our policies effectively mitigated impacts on system performance during our simulated scenarios.Figure 3 highlights the performance of the informed policies relative to 'status quo' by depicting this policy performance as bars and the relative change of the policy performance as lollipops.Therefore, a downward-facing lollipop shows a decrease (i.e.improvement) in objective performance relative to 'status quo' .
The scenarios applied to our case study reveal that external stressors can have widely varying impacts on policy performance.Recall the tendency from section 4.1 that the two policies that attempt only water reductions tend to strongly emphasize water-related objectives at the severe cost of all other objectives.The extreme load/climate scenario illustrates the severity of this tendency.By introducing a high consumption penalty, some generators discharge significant volumes of elevated-temperature water beyond legal limits posing threats to riverine ecology.Conversely, the Figure 2. Policy performance over several scenarios.Objectives, scenarios, and policies are depicted as rows, columns, and colors, respectively.We introduce the discharge violations objective defined as the amount and temperature of water discharged beyond the legal limit as well as the energy not supplied by the system (i.e.reliability) as additional system objectives.Objective performance is impacted by the scenario regardless of the policy.'high withdrawal penalty policy' impacts system reliability as the cost of withdrawn water outweighs the cost of not serving demand.However, the nuclear outage scenario illustrates that these two policies actually improve emissions while not impacting reliability or discharge violations.
A policy focused only on emissions reduction has benefits for the various water-related objectives of the system across all simulated scenarios.We see the emission-focused policy ('high-emissions penalty') only marginally (less than 6%) increases cost across all scenarios.However, for this cost increase, it is able to unanimously improve status quo water withdrawal, water consumption, and emissions performance.It even offers some benefits to reducing thermal discharge violations in the extreme load/climate scenario.Importantly, we see that the reliability of the system is maintained with an emissionsonly-focused policy.In the next section, we discuss how our holistic water and emissions policy is able to further improve on the emissions-only policy.
Figure 3. Policy performance relative to status quo operations.'Status quo' policy is depicted as bars.Downward lollipops portray an improvement in policy performance relative to 'status quo' policy.Through water-and emission-informed policy, the impacts of various climatological and operational stressors can be effectively mitigated.

The benefits of holistic policy considerations
Life-threatening impacts on fish populations due to thermal discharge violations from power plants (Logan et al 2021, Barnthouse and Coutant 2022) are avoidable through well-created and enforced energy policies.However, the physical grid reliability can be threatened by policies that completely eliminate such violations (e.g.'high water withdrawal penalty').Meanwhile, other policies offer only some reductions to these thermal discharge violations (e.g.'high emissions penalty').
We propose a holistic 'water-emissions' policy that capitalizes on some of the strengths of the 'high emissions penalty' policy while augmenting the ability to operate the system with no discharge violations.Specifically, this holistic policy leads to no thermal discharge violations while yielding improvements in system water withdrawal, water consumption, and emissions without major increases in operational cost.These benefits to emissions also hold true when looking at the system-wide average emission rates (SI figure 4).Additionally, this policy's objective performance is preserved across climatological and operational stressors.
Our 'water-emissions' policy also has enforcement benefits.The proposed policies are incorporated directly into the existing operational problem of OPF without adding substantial complexity.Such direct incorporation avoids the need for enforcement that is dependent on human judgment.For example, currently, many existing laws prevent discharge temperature violations.However, such laws are rarely enforced (Micha 2014, Liu et al 2017).In the avoid temperature violation scenario, we simulate a reality in which such laws are enforced and generators reduce capacity instead of allowing temperature violations.
Even under such scenarios, our policy still offers benefits to reducing water withdrawal and emissions.

Conclusions
Through simulation applied to a case study, we discover fundamental interactions among system operational costs, water withdrawal, water consumption, emissions, and reliability.We pose alternative water-and emissions-informed policies for operating energy systems such that impacts associated with extreme operational and climatological system stressors can be effectively mitigated.However, making such policies without considering the coupling of systems can lead to severe degradation of system performance.Instead, we propose a multi-objective policy search framework that allows policymakers to create holistic policies that quantitatively reflect their specific preferences for system operation and can be directly incorporated into existing power system operations.
In the future, the objective policy preference discussed in this work could be expressed in real-time by system operators.For example, frameworks and figures like the ones implemented in this work could be implemented in a graphical user interface to act as an assistive tool for system operators.Such interactive multi-objective policy visualization tools have been implemented in other water resources engineering problems (Smith et al 2017, Raseman et al 2019).
Although this work presents an effective policy that can be directly incorporated into existing power system operations, further research efforts are needed regarding the equitability of such policies.Although the resulting cost increases were marginal, such increases can add up over time, and thus further efforts are needed to determine how these costs could be equitably split to answer the 'who pays?' question of energy policies.Framing such future work as a sensitivity analysis could also better answer the question of what policy choices give you the most benefit for cost increase.
This work also showed how operational policies optimized to an average week had the same efficacy on several more stressful scenarios.However, further efforts could take the view of ensuring robust policies that are designed to be effective across many possible states of the world (Kasprzyk et al 2013).Similarly, future research efforts must examine how our proposed policy performs on longer timescales.Such an analysis could be done using a full production cost modeling approach to consider policy impacts on energy market processes such as unit commitment as well as seasonal storage/demand fluctuations.One advantage of the policies proposed in this work is that they are considered in the objective function and thus could easily be added to existing production cost modeling frameworks.
The important choice of how to operate the energy grid leads to cascading impacts that are felt far beyond the energy sector.Neglecting such interactions during operation can have detrimental externalities.While it may be tempting to consider such externalities as a hard constraint on system operations, we have seen that when energy grids are pushed to their limits, these once-hard constraints seem like ideals.Multi-objective operational policy frameworks, such as the one proposed in this study, demonstrate the ability to more honestly and consistently represent the connectedness of the modern energy system.Therefore, if we are to make effective policies for modern coupled energy systems, systems that are inherently multi-objective, we must embrace multi-objective energy policies.
• F per Illinois regulation (Illinois Environmental Protection Agency 2005, Meyer and Wernau 2012).Maximum withdrawal and consumption rates are based on national estimates (Macknick et al 2012) as well as the regional estimates in Kravits et al (2022b).A value of lost load of $3300 MWh is used given historic Midcontinent Independent System Operator (MISO) values (Cook 2021).

Table 1 .
Generated policies.Each policy captures varying degrees of economic, emissions, and water considerations.

Table 2 .
Scenarios.The load and climatological stressors are set to average levels for these scenarios (i.e.outages are the only differences, with all other factors remaining constant).
a For example, due to either maintenance or a water-related outage Freedman (2012), Krier (2012).b Madden N, Lewis A and Davis M 2013 Thermal effluent from the power sector: an analysis of once-through cooling system impacts on surface water temperature Environ.Res.Lett.In hot water: clean water act provisional variances and their relationship to the impact of heat waves and droughts on the supply and demand of electricity Chi.-KentJ. Environ.Energy Law 4 1-34 (available at: https:// studentorgs.kentlaw.iit.edu/ckjeel/v4i1-2013-2014-4-micha/)Mohseni O, Stefan H G and Erickson T R 1998 A nonlinear regression model for weekly stream temperatures Water Resour.Res.34 2685-92 Mu M, Zhang Z, Cai X and Tang Q 2020 A water-electricity nexus model to analyze thermoelectricity supply reliability under environmental regulations and economic penalties during drought events Environ.Model.Softw.123 104514 Myhre R 2002 Water and sustainability (volume 3): U.S. water consumption for power production-the next half century Technical Report 1006786, EPRI Palo Alto, CA (available at: www.circleofblue.org/wp-content/uploads/2010/08/EPRI-Volume-3.pdf)Pacsi A P, Alhajeri N S, Webster M D, Webber M E and Allen D T 2013 Changing the spatial location of electricity generation to increase water availability in areas with drought: a feasibility study and quantification of air quality impacts in Texas Environ.Res.Lett.8 035029 Pfenninger S, Hawkes A and Keirstead J 2014 Energy systems modeling for twenty-first century energy challenges Renew.Sustain.Energy Rev. 33 74-86 Poumadére M, Mays C, Le Mer S and Blong R 2005 The 2003 heat wave in France: dangerous climate change here and now Risk Anal. 25 1483-94 Raseman W J, Jacobson J and Kasprzyk J R 2019 Parasol: an open source, interactive parallel coordinates library for multiobjective decision making Environ.Model.Softw.116 153-63 Reuters Staff 2011 Drought adds to 2012 Texas power supply worry Reuters Rhodes J 2017 The Conversation (available at: https:// theconversation.com/the-old-dirty-creaky-us-electricgrid-would-cost-5-trillion-to-replace-where-shouldinfrastructure-spending-go-68290)Rutberg M J, Delgado A, Herzog H J and Ghoniem A F 2011 A system-level generic model of water use at power plants and its application to regional water use estimation Advances in Aerospace Technology; Energy Water Nexus; Globalization of Engineering; Posters vol 1 (ASMEDC) Sanders K T, Blackhurst M F, King C W and Webber M E 2014 The impact of water use fees on dispatching and water requirements for water-cooled power plants in Texas Environ.Sci.Technol.48 7128-34 Sengupta M, Xie Y, Lopez A, Habte A, Maclaurin G and Shelby J 2018 The National Solar Radiation Data Base (NSRDB) Renew.Sustain.Energy Rev. 89 51-60 Smith R, Kasprzyk J and Dilling L 2017 Participatory framework for assessment and improvement of tools (ParFAIT): increasing the impact and relevance of water management decision support research Environ.Model.Softw.95 432-46 Stone B J et al 2023 How blackouts during heat waves amplify mortality and morbidity risk Environ.Sci.Technol.57 8245-55 van Vliet M T H, Wiberg D, Leduc S and Riahi K 2016 Power-generation system vulnerability and adaptation to changes in climate and water resources Nat.Clim.Change 6 375-80 Zatarain Salazar J, Reed P M, Herman J D, Giuliani M and Castelletti A 2016 A diagnostic assessment of evolutionary algorithms for multi-objective surface water reservoir control Adv.Water Resour.92 172-85 Zhai H and Rubin E S 2010 Performance and cost of wet and dry cooling systems for pulverized coal power plants with and without carbon capture and storage Energy Policy 38 5653-60