The climate-energy nexus: a critical review of power grid components, extreme weather, and adaptation measures

A recent example of heatwave impacts on demand is during the 2022 September heatwave in California, when the highest electricity peak demand on record for the California Independent System Operator (CAISO) was reached on 6th September, 2022, at 51 426 MW. CAISO reported that the Western Interconnection set a new peak demand that day at 164.6 GW (California Independent System Operator 2022a). While this event did not result in rolling blackouts, DR was heavily utilized. In September, DR was responsible for a demand decrease of 1882 MW. Regarding DR, Mayhorn et al (2015) has dubbed four terms to describe how building loads can respond to the grid provided in table 1. Lastly, Battery Energy Storage Systems (BESS) could provide grid stability during heatwaves, as BESS provided 6.8% of supply during the September heatwave in California (California Independent System Operator 2022a).

For transmission and distribution systems, Sathaye et al (2012) conducted a study for the California Energy Commission and found that transmission capacity could potentially be reduced 7%–8% with an increase in air temperatures, specifically a 5 °C increase on an August day. While Bartos et al (2016) investigated 121 planning areas in the U.S. during peak summer demand and found summer reductions of transmission capacity ranging from 1.9%–5.8% on average from 2040–2060, compared to 1990–2010, depending on the GCM and emissions scenario, due to rises in ambient air temperature.

Sathaye et al (2013) found a decrease in transformer capacity up to 2.7% induced by increases in ambient temperatures, which would require a projected 3.5% more generation needed to overcome losses.

The relationship between environmental variables, including ambient temperatures, will vary across regions and substations size, Swift et al (2001), Li et al (2005), Sathaye et al (2013)work to quantify the impacts of ambient temperatures on transformer capacity and lifespan, as Li et al (2005) found a decrease in transformer capacity equating to 0.7% for every 1 °C increase in ambient temperature. However, this varies, given the allowable HST limit and the cooling equipment installed.

In Steinberg et al (2020), generation capacity and load are investigated through 2050, primarily considering drought and temperature deviations. Intuitively, demand increased with increases in temperature for the CONUS, with an average demand increase of 1%–2% for every 1 °C increase in air temperature with regional differences, the magnitude of change depended on the chosen climate scenario.

Heatwaves impact generation primarily through efficiency losses. For the Central and Eastern U.S., thermal, fossil-fuelled and nuclear power plants that supply 11% of electric power production for the entire U.S. is studied in Vliet van et al (2012). Results found that plant reductions of 25% occurred on 29–30 days, 50% on 15 days, and 90% for one day. Through the lens of compliance contingencies for thermal plant cooling, such as water temperature violations and stream flow waivers, an average reduction in generating capacity of 2%–3% is predicted for thermoelectric plants by the 2060s with streamflow waivers. However, reductions up to 12% are possible if waivers are not granted (Liu et al 2017a). While Miara et al (2013) suggests a 2.5% decrease in capacity in summer and highlights the need for regional studies to better understand thermal plants' environmental, hydrological, and energy impacts.

Outside of thermal generation, solar energy production decreases with rising temperatures dependent on the material and type of solar plant. This occurs due to panels generating less voltage, which decreases efficiency in producing electricity. For Concentrated Solar Power (CSP) and Solar Photovoltaic (PV), Crook et al (2011) used GCMs to study solar production worldwide. In the U.S., CSP showed an increase in the Northwest and a decrease in the Southwest, while PV output is predicted to experience a minor decrease in the Western U.S. However, more regional studies should be performed.

To adjust to more frequent heatwaves in the future Alawad et al (2021) considers heatwaves from an operational standpoint under conditions from the 2011 Texas Heatwave, a 40-day period having a consecutive temperature above 38 °C.

Other studies such as Choobineh et al (2016) investigate how the optimal operation of a distribution system can change during a heatwave. Findings suggest that dispatch cost increases with increased temperatures; further, the dispatch model purchases less power from the grid despite an increase in load. This occurs under heatwaves due to the effect of temperature on the overhead line and gives prominence to the argument that even with adequate capacity, the system becomes strained due to energy transport constraints (generation limitations, equipment inefficiencies, increased prices, and line losses) under high temperatures.

A large-scale example occurred in 1998 in Eastern Canada and the Northeast U.S. when freezing rain destroyed approximately 1000 transmission towers and 30 000 utility poles, causing power outages and disrupting transportation systems (Jeong et al 2019). Alternative methods are used by utilities in Canada to defrost overloaded lines. Quebec, Canada, has the only high voltage D.C. not used for transmission but for de-icing multiple A.C. power lines, known as the Levis De-Icer (Kirby et al 2008).

In December 2022, National Grid, an electricity and gas provider in Buffalo, New York, experienced blackouts that impacted thousands of customers due to substations going offline due to frozen transformers, responsible for stepping down the voltage to the distribution level (National Grid 2022), creating the need for greater weatherization procedures for transmission and distribution systems.

In 2014, a polar vortex impacting the Eastern interconnect set a new winter peak demand for the CONUS at 140 510 MW in the Pennsylvania-New Jersey-Maryland Interconnection (PJM) territory. PJM is a regional transmission organization that governs electricity operations for 13 states and the District of Columbia. During the 2014 polar vortex, PJM implemented DR three times, one DR event lasting over four hours, to attempt to reach an equilibrium between supply and demand (Wang et al 2017). As the outage rate for generators, primarily for thermal generation, in the PJM territory was 22% compared to the previous average of 7% (Interconnection 2014).

The 2021 Texas Winter Freeze is a recent example of how extreme winter weather can impact power generation. One prominent cause of the cascading failures was that approximately 40% of natural gas production was unavailable. Freezing gas wells and pipelines that transport natural gas were two significant causes relating to compressor stations experiencing power outages. Compressor stations maintain natural gas flow and pressure from the pipelines. Texas is more prone to warm periods. Thus, most power stations are not in buildings, which traditionally helps minimize overheating but exposes them to freezing winter conditions (Busby et al 2021). It is important to note that other energy resources, such as coal, nuclear, and wind, also struggled to generate electricity during this time, primarily due to inadequate weatherization.

Relating to natural gas well freezing or 'freeze-offs,' there are typically two ways this occurs:

• 1.  
  Above-ground and wellhead structures may be vulnerable to simple freezing, as the liquid turns to ice and blocks the wellhead or outward-bound pipes, although buried pipelines could be susceptible to this with extremely low ground temperatures.
• 2.  
  Combined high-pressure and low-temperature environments turn liquid and liquifiable hydrocarbons into hydrate masses that block the line.

System design criteria typically use 30 years or, in some cases, 100-year historical data for design assumptions. The 2021 Texas Freeze surpassed historical weather, with record low temperatures from the past 150 years. Thus, the system design was not meant to withstand such cold weather conditions (Smead 2021). Moving forward, it is essential for traditionally warm climates, especially those in the Southern U.S., to take preventative measures even if historical freezing temperatures are not typical. In the case of Texas, previous winter storms have affected the state, including those in 1989, 2003, and 2011.

Regarding the Texas 2021 freeze, emergency load shedding was implemented to avoid a system-wide blackout. During the three coldest days of the winter freeze, the electricity wholesale market reached a maximum cost of $9000/MWh. The wholesale cost before the freeze ranged from $20/MWh–$40/MWh. Concurrently, natural gas reached a maximum wholesale cost of $117/MWh, with prior costs ranging from $0.88–$1.17/MWh)

Future concerns for power infrastructure include findings from Jeong et al (2019), Li (2023)who predict that ice loads pose a threat to infrastructure in Northern North America, but Southern North America and the Northeast coast may experience a decrease in ice loads. This highlights the spatial complexity of climatic change and its impacts on critical infrastructure.

With increased building electrification, the electrical load will likely increase during severe winter weather. This creates a question regarding dual-season/winter peaking in electrical grid operations. Traditionally, in the U.S., the annual peak is in the summer, and heating was often gas-powered. For instance, Keskar et al (2023) suggests that areas within PJM, such as West Virginia, Virginia, and Kentucky, may exhibit winter peaking behaviors in the future. Winter peaking could be particularly challenging for investor-owned utilities since most must submit integrated plans to the state's public utility commission (Keskar et al 2023). Additionally, battery storage systems may also struggle during extremely cold temperatures as they hold less charge and drain more quickly (Lu et al 2022). Although winter peaking often occurs in Europe, the implications for grid operations with a transition from winter to summer peaking or dual peaking remain unknown.

To operate the grid, Keskar et al (2023) suggests seven policy recommendations for grid planners and operators to ensure grid reliability throughout the winter weather. These include increasing transmission capacity, resource adequacy planning for the relationship between temperature and generator failures, implementing similar policies used by winter peaking regions, and determining demand-side management techniques suitable for a winter peak.

Approximately one-third of U.S. thermoelectric capacity reports a primary water source as a lake or reservoir. Reservoirs were the reported water source for approximately 25% of curtailments between 2000–2015. Such curtailments are most commonly caused by a lack of cooling water, high intake temperatures, or high discharge temperatures.

Regarding hydroelectric generation, Washington, California, and Oregon contribute the highest percentage of hydroelectric power to the grid, specifically the Western Interconnect. Droughts in 2021 presented the lowest hydropower generation in the Western Interconnect since 2001. This results in generation levels nearly 16% below the average for the 21st century. Beyond the Western Interconnect, total annual precipitation strongly correlates with total annual generation from hydropower plants in five out of eight climate regions in the U.S. (Turner et al 2022).

In Cohen et al (2014), a study investigating water rights availability, power system operations, and electric power growth finds surface water availability had the most significant decrease in the Southwest, with regional differences in water withdrawal resources. Most of the water withdrawal attributed to existing nuclear and coal-fired plants in the East, where water is less constrained. The report also highlights how averaged projections may be misleading regarding power system sensitivity, as individual systems often experience stress during weather events, not the average.

Findings from North American Electric Reliability Corporation (2022) highlight Resource Adequacy and Risk issues across the U.S., including decreases in energy output from hydropower generation due to drought and low snowpack in the West, drought, and heat events that could heavily influence peak demand in Texas and the Midcontinent Independent System Operator capacity limitations resulting in energy emergencies at peak demand. Accordingly, Bartos and Chester (2015) found that renewable power generation is generally less vulnerable to the impacts of climate change, with the most significant constraints coming from an extreme drought event. By the mid-century, a ten-year drought could reduce summertime capacity by 6.6–8.0 GW, 3%–4% of the Western Interconnection.

Due to droughts' impacts on generation, it is vital to better understand the balance between supply and demand during precipitation extremes. In Jornada and Leon (2016), electric generation decision-making using a multiobjective function to minimize cost and water withdrawal is studied, while Kumar et al (2022) incorporates water into the economic dispatch problem. Further, Wu et al (2010), Hadji et al (2015), Kravits et al (2022a, 2022b) incorporate water metrics in the Optimal Power Flow (OPF) problem. A common finding suggests that implementing water constraints as part of the OPF may result in more proactive rather than reactive operations. Such advancements in operational frameworks for periods with water stresses have the potential to aid in identifying system vulnerabilities and intercomparison of vulnerabilities across the network (Kravits et al 2022a).

In contrast, an electric grid dispatch model provides insight into years of increased stream flow followed by years of drought. This model shows that years with increased precipitation only sometimes equate to increased hydropower generation due to spill events. Overall, they found a slight decrease or no change in hydropower generation but greater temporal variability (Tarroja et al 2016). Precipitation may also impact production costs and utility profit margins. Moving forward, dose-response functions should be further explored as they are a type of regressions model that can quantify the physical response of a system component in the system to its exposure to climate stress (Dumas et al 2019).

Extreme winds predominately impact transmission and distribution systems through structural damage. To study structural impacts on the power system, lattice power towers are modeled to capture buckling and post-buckling effects, joint slips, and power tower failures due to extreme winds. These events can be catastrophic given that power towers carry overhead power lines, generally high-voltage transmission lines. To model the lattice transmission towers, finite element modeling is paired with a Monte Carlo simulation for uncertainties of failures. Findings conclude that buckling from extreme wind, on average, decreases load-bearing by 30%. Additionally, findings show that elements experience partial failures prior to complete failure. These findings could be used to assess vulnerabilities and aid in recovery strategies (Mohammadi et al 2020).

Specific to distribution and low-voltage power lines is the fragility of wooden poles to extreme wind speeds, especially as power infrastructure is aging, with most wooden poles being over 30 years old (Pansini 2020). One way to gauge the impacts of wind on power infrastructure is fragility curves and damage functions. Fragility curves provide a probability of system or component failure in response to exposure to extreme wind. Damage functions relate the intensity of wind to the mean damage ratio. Numerous methods exist to develop these curves, including judgmental, empirical, analytical, and hybrid (Dumas et al 2019).

Time-dependent fragility curves are studied in Shafieezadeh et al (2014) concerning wind speeds, using regulated wooden pole types. Findings conclude that focusing on examining poles 25 years and older in wind-prone areas was consistently determined as the most high-risk parameter across pole types. This promotes using such methods in risk-informed decision-making for maintenance, inspections, and replacements. Further, the study found a failure probability of one for transmission lines at winds of 65 $\frac{m}{s}$ and poles at 100 $\frac{m}{s}$.

Due to the cut-in and cut-out speeds, wind speed and direction impact wind power production. In Pryor and Barthelmie (2010), a review of climate change impacts on the wind energy industry was conducted, highlighting that many studies focus on wind availability, but other climatic factors such as extreme winter storms, icing the wind turbine blades, high and low temperatures, impacting physical components, the ability to access remote wind farms, and extreme wind gust that result in curtailing the wind turbine, need to be understood at a deeper level.

While Sailor et al (2008) used two GCMs to estimate wind energy potential under climate change. The results showed a considerable disagreement between the models, with one model suggesting minimal impacts on wind resources, while the second model suggests mean wind reductions of 10%–15%.

Similarly, Karnauskas et al (2018) finds that when using RCP 4.5–8.5 with 10 GCMs, wind power decreases 8%–10% for the CONUS, yet increases in the Eastern U.S. by 2050. In contrast, Pryor et al (2012) suggest that wind power generation will decrease in the Western, Northeast, and Northwest U.S. but increase in parts of Texas and Kansas. Overall, extreme winds' impact need to be understood further, but the CONUS should expect a decrease in generation in some areas.

This agrees with Pryor et al (2023) who studies the change in wind power capacity factors and projected power under climate change, using dynamic down scaling techniques. Findings suggest that projected power remains largely unchanged East of the rocky mountains in North America. However, five wind energy facilities show significant decreases, three in the midwest, one in the Northern Greater Plains, and one in the Northeast. Further, Jung and Schindler (2019) finds a decrease in average near-surface wind especially in U.S. West, using a suite of models and RCP 8.5.

Offshore wind, the U.S. has a goal to have 30 GW of offshore wind capacity by 2030 win (2023). Yet, limited research is available regarding the impacts of climate change on off-shore wind in the U.S. Fatigue damage and structural saftey of off-shore wind turbines were found to be sensitive to conditions such as wind speed and wave height in future climate scenarios, while non-structural components were found to be more robust Wilkie and Galasso (2020). In the United Kingdom, (Abdelaziz et al 2024) studies off-shore wind potential during times of low wind speeds. While (James et al 2023) highlights how climate variables such as wind speed, wave height, and sea ice melt will impact off-shore wind in the future.

Approximately 80% of past vegetation-related issues for power systems result from fallen trees or branches into transmission and distribution systems. These incidents are often caused by trees that are not part of the electric utility's right-of-way (Russell et al 2012). However, vegetation management is a challenging and expensive task (Wischkaemper et al 2008). In the U.S., it is common for utilities to trim approximately 12 feet of clearance away from distribution lines. However, a case study in the Southeastern U.S. suggests that vegetation debris from up to 100 feet from power lines could evoke damage to the system (Guikema et al 2006).

The most destructive and deadliest wildfire in the history of California, the 2018 Camp Fire, was caused by faulty power line components. The Camp Fire resulted in 85 deaths, 18 804 structures destroyed, and $400 billion in total cost (insured losses, debris removal, and victim aid). Between 2015 and 2020, electric power caused an average of 10% of wildfires in California while being responsible for an average of 33% of acres burned and 70% of damages ($17.5 billion) (Cal-Fire 2020). Between 2000-2016 wildfires cost utilities in California more than $700 million in damages (Dale et al 2018).

Wildfires can also impact power generation, as in 2020 CAISO experienced an approximately 30% reduction in solar production due to wildfire smoke, which contains airborne particles with a diameter of 2.5 micrometers (PM2.5) that reduces the exposure of sunlight to solar arrays, reducing the electricity produced (York 2020).

Current approaches utilities take to operate the grid during wildfires include Public Safety Power Shut-Offs (PSPS). PSPS turns off or de-energizes the power to lines in areas with high wildfire risk. De-energization effectively reduces fire risk by preventing fault events that could ignite a wildfire (California Public Utilities Commission 2023). However, PSPS can result in prolonged blackouts, having adverse health, economic, and societal impacts, particularly for venerable socio-economic communities, often in rural WUI areas (Sotolongo et al 2020). Approximately one-third of wildfire-related outages were due to PSPS.

Another threat to the wildfire-power infrastructure nexus is the expansion of the Wildland-Urban Interface (WUI), where the built environment imposes the natural environment. It is estimated that around 50 million residential buildings are in the WUI, and approximately one million additional homes enter the WUI every three years (Burke et al 2021). This means distribution systems and transmission lines have emerged in the WUI. Distribution lines impose a threat as they are three times more likely to ignite a wildfire than transmission lines on a per-mile basis (Pacific Gas and Electric Company 2019).

A late-summer wildfire season in the Western U.S., is projected to pose a reliability risk to the bulk power system (North American Electric Reliability Corporation 2022). Further, in Cai et al (2023), the years 2017–2020 are studied relating to solar power generation and solar power forecasting errors in CAISO during wildfires. In this study, 2019 was used as the baseline year, as it was the year with the smallest wildfires. Compared to 2019, each year studied experienced reduced solar power production and increased forecasting errors during a wildfire event. Regions with a high penetration of solar energy should consider how wildfire events may impact the generation mix and be prepared to use backup generation.

To address these issues, recent studies have aimed to balance the effect of blackouts with wildfire risk (Rhodes et al 2021, Rhodes and Roald 2022, Umunnakwe et al 2022). A dispatch model considered how preventive wildfire risk measures impact both wildfire risk and power systems reliability in Rhodes et al (2021). The optimization problem aims to reduce the risk of wildfires from power lines and meet the maximum load and was successful in serving more load at a lower risk.

Building upon Rhodes et al (2021), Rhodes and Roald (2022)implemented a rolling horizon co-optimization problem, optimizing both power shut-offs considering wildfire risk and post-wildfire restoration. Results suggest that using a longer forecast horizon allows for a better understanding of the impacts of line de-energization. However, a longer forecast horizon typically results in more significant forecast errors. The larger the restoration budget, which represents the available resources to inspect and restore power lines, the greater the number of lines de-energized and more quickly restored. A higher vulnerability threshold increases system redundancy, increasing wildfire risk and power line operational security. While Umunnakwe et al (2022) found spatial resolution to be an important factor in modeling PSPS, as forced outages due to de-energization were decreased with the use of more granular spatial models.

Hurricane impacts on the power system have been widespread in recent years. Hurricane Maria devastatingly destroyed Puerto Rico's grid in 2017, which took approximately 11 months to completely restore, marking the most continuous blackout in U.S. history, with only 65% of the grid being restored by the end of January 2018, as 80% of utility poles and a significant portion of transmission lines were destroyed (United States Department of Energy 2018a). The following year, in 2018, Hurricane Michael became the first category five hurricane to make landfall in the CONUS, resulting in outages spanning six states and 1.7 million customers (United States Department of Energy 2018b).

In the case of Hurricane Maria, wind force vs. deforestation and rainfall vs. landslide were shown to be the most significant in terms of coupled hurricane factors through correlation factor analyses and modeled structural failures (Chen et al 2021).

For power generation, natural gas systems are a crucial issue when considering hurricanes. During Hurricane Katrina, approximately 88% of daily gas production was shut-in (i.e. the valves of natural gas wells are closed to stop production), and 20% of natural gas capacity remained shut-in through 2005. Similarly, 80% of natural gas was shut-in during Hurricane Rita. Considering coal, the Mississippi River was used for coal transport, which was interrupted for multiple weeks after Hurricane Katrina (Gil and McCalley 2011).

For Hurricane Sandy, approximately 32 000 consumers were affected due to the lack of gas distribution, highlighting large-scale disruptions caused by hurricanes and the loss of transmission or gas pipeline, which could quickly distribute in the system and lead to cascading failures. The interdependency between the electric power and natural gas systems has increased due to the gas-operated distributed generation and electric gas compressors (Davidson et al 2003, He et al 2018, Waseem and Manshadi 2020). Additionally, other times of power generation, such as ground-mounted solar power, could be impacted by flooding due to hurricanes.

In Yuan et al (2016), numerous optimization models were tested to effectively budget resources for hardening against hurricanes, a resilient distribution network planning problem. Findings suggest a two-stage robust optimization problem adequately includes the uncertainty of such natural disasters, extending the use of traditional N-K interdiction models. N-K models model the number of K power components that, if failed, would result in a worst-case power disruption, as N represents the number of power lines in the system.

Similar to wildfires, distribution systems suffer more from hurricane damage than transmission systems, given the materials used in distribution system infrastructure and design (Davidson et al 2003). However, numerous studies have identified vulnerabilities in the transmission system, such as in Liu and Singh (2011), which uses a load model, generation system, and transmission network to explore the nonlinear relationship of hurricane parameters and increment multipliers failure rates of the transmission lines. The models effectively evaluated the temporal and spatial impacts of the power system against hurricanes and found that power line damages are a primary reason for power outages during hurricanes.

Power system operations during hurricanes traditionally entail immediate solutions to reduce impacts as they occur, especially if the transmission and distribution system is severely damaged and the ability to transport power is limited. A generation redispatch strategy to determine the optimal generation level of operable generators during a hurricane is studied in Abdelmalak and Benidris (2022), which minimizes load curtailments during component failures by approximately 40% by redispatching power supply after subsequent failures.

Additionally, microgrids offer a novel approach to operations during hurricanes. In Mitra and Ranade (2007), Chanda and Srivastava (2016), Chen et al (2017), Liu et al (2017b), methods proving microgrids enhance network resilience and offer a self-supply of electricity during emergencies are explored. While Amirioun et al (2018) runs a resilience-constrained OPF to minimize load curtailment using microgrids during floods. In Watson and Etemadi (2020), microgrid resilience is modeled using Hurricane Harvey as a case study, in which outages were significantly reduced.

Although not in the CONUS, the summer of 2010 in Russia is a prime example of compounding events that mirror similar events in the U.S. West, but at a larger scale. Russia experienced below average precipitation that resulted in a summer drought, which coincided with an unprecedented heatwave. The drought, compounded with the heatwave, led to widespread wildfires. These wildfires increased the death-toll due to air pollution across major cities, as more than 10 000 Moscow citizens died during these events (Barriopedro et al 2011, Zscheischler et al 2018). Considering the CONUS, 2020 had widespread droughts combined with abnormal ambient temperatures throughout the West and central states. This additionally intensified Western wildfires. On a national scale, December 2020 the largest area of land since 2012 was experiencing extreme drought conditions, which was emphasized by record heat in parts of the PNW (dro 2023). Lastly, when considering the top 25 extreme weather events for the CONUS, extreme winds are often coupled with blizzards, wildfires, and severe storms (U.S. Department of Commerce 2023).

The impacts of wind and solar in Texas are studied in Craig et al (2020), and findings suggest that wind power production may increase due to changes in wind speed, simultaneously demand will increase due to an increase in ambient temperatures, creating issues with supply and demand. While Tavakol et al (2020) studies the impacts of hot, dry, and windy events, which are more likely to occur during a heatwave and drought. Results show that hot, dry, and windy events increase from Southwest Kansas into Northwest Texas, which could also create supply and demand issues if wind speeds are too high. Turner et al (2019) studies the impacts of increased demand due to ambient temperature and drought, and finds the loss of load to double by 2035 due to demand increase and water scarcity. Additionally, the study found that while power outages were more frequent, they were also shorter in duration, offering adaptation measures the ability to improve power system performance in the years to come. Considering thermal power plants under a co-occurring heatwave and drought, Lubega and Stillwell (2018) creates operational rules that determine the minimal number and magnitude of thermal variance from a power grid perspective, which allows power plant and power grid operators to input information. With increased temperatures and decreased precipitation, thermal variances increase, highlighting the importance of contingency planning. While, Cohen et al (2022) links hydrology models with a grid planning model to test future power grid scenarios, including an increased demand with reduced hydropower due to drought, with multiple RCPs. While results varied for each RCP, generation and transmission capacity increased, along with operating costs. Lastly, Gurgel et al (2024) studies the gap in supply and demand in the Central U.S. and finds the most significant supply gap at 21% to occur in the Lakes-Mid Atlantic region as a result of increased demand and decreases in thermal generation due to heat and drought.

Considering approaches for adapting to heat waves, past experiences such as the September 2022 heatwave in California offer invaluable insight. CAISO released adaptation strategies that supported the grid throughout the heat wave. These include but are not limited to an increase in resource adequacy, including 3500 MW of battery storage, state programs to provide non-market resources (including emergency/flex alerts to customers), emergency assistance that allows for energy flow to and from the CAISO grid, this demonstrates the importance of resource-sharing and a connected grid (California Independent System Operator 2022a). Areas identified to be improved operations coordination between day-ahead and real-time battery dispatch, a software upgrade related to how the market chooses or curtails low-priority exports under unconventional conditions, and changing the calculations used in the Western Energy Imbalance Market resource sufficiency test, which includes power flow feasibility, balancing supply and demand, flexible generation ramping, and bid-in range capacity (California Independent System Operator 2022b).

To increase awareness of the impacts of heatwaves on the grid, a better understanding of the variability of both load and power transport regionally is needed, as well as operational strategies that can minimize rolling blackouts. One solution could be increasing renewable energy generation capabilities with more advanced operation strategies and the DERs to aid in curtailment measures. This may require advances in inverter technology, more efficient scheduling for BESS, and advances in operational software. Thus, more research should be performed.

Adaptations for extreme winter weather should balance power system resiliency, reliability, and crew dispatch safety. Protecting the distribution and transmission system during extreme winter weather is often a manual endeavor in the U.S., which includes crews removing the ice or salting the power lines. Predictive models such as those studied in Zarnani et al (2012) offer solutions for crew dispatch and preparedness for such tasks, yet aerial vehicles also hold the potential to perform such tasks.

Weatherizing power generation plants plays a vital role in maintaining power production. Weatherization may include line heating, antifreeze injectors, and enclosures to trap heat in the winter and release heat in the summer for natural gas infrastructure and equipping wind turbines with antifreeze coatings and heating elements. However, retrofitting existing natural gas plants is expected to be a costly endeavor (Smead 2021). There is progress in winter infrastructure weatherization. In February 2023, FERC approved two standards, 'Extreme Cold Weather Reliability' and 'Directing Modification Of Reliability', which target power generators' retrofitting and construction process during winter weather. Weatherization alone will not be enough if fuel, particularly natural gas, is not available. Resource adequacy measures must consider the relationship between generator outages, generator maintenance, and temperature, thus reconsidering annual maintenance measures in winter for traditionally warm climates such as Texas.

Additionally, regions such as Alaska may experience an increase in cost between 10% and 20% by 2030 to maintain and expand public infrastructure due to climate change (Larsen et al 2008). Alaska is particularly vulnerable to climate change; as a traditionally cold climate, it is warming two to three times the average global rate, putting infrastructure at risk due to permafrost thawing and floods increasing. Thus, system hardening, modernization, and weatherization are vital in proactive strategies. However, more research on equitably adapting the power grid in rural communities and tribal lands should be further studied and considered.

Adaptive measures for both thermal generators and hydropower plants are not straightforward. For thermal power plants, new plants should avoid once-through cooling systems and utilize water reuse strategies, while existing plants should be retrofitted for reuse strategies. Retrofitting thermal and hydropower plants includes adjusting intake level requirements to be lower. Other adaptation strategies for both generation types include increasing monitoring of reservoirs and stream flows, increasing power production efficiency, and increasing water storage capacity, aquifer storage, and groundwater recovery without causing harmful environmental impacts. Further, utilities should create contingency plans for times of extreme drought. Droughts can cause interdependence between regions as power may need to be exported, and scenarios creating water restrictions for users have created opportunities for better operations within the water-power nexus.

For improved operations, future studies should link OPF programs with physics-based models, hydrologic/hydraulic models, RCMs, or system loadings and include more exogenous parameters across multiple time steps and regions. Exogenous parameters include any parameter that may change depending on the system or parameters the system operator can not control. These could include optimization weights, generator water consumption, rate of generator water withdrawal, and system loadings, many of which would change per timestamp (Kravits et al 2022b).

Overall, thermal and hydropower are significant contributors to the U.S. power grid, with hydropower aiding in carbon migration efforts as the largest percentage of renewable energy generation. Thus, operating the power grid during droughts, without outages or water use constraints for the general public, is an area of increasing research interest.

Adaption strategies for wind alone predominately consist of frequent inspections and system hardening measures, as presented in table 2. Infrastructure materials (wood or steel poles) and bracing strategies (guy wires) should be deployed for wind-prone areas. As Li (2023) highlights the importance of the structural design of wind turbines and the structural reliability of wind power poles from simulated results using 15 RCMs and RCP8.5. Adaption measures should consider construction cost and structural safety, as structural reliability is projected to decrease in many areas across North America. Alternatively, retrofitting the current system to include underground power lines would prevent wind-driven damages and routine vegetation management procedures.

Modernization of the power grid, such as increased sensor technology to detect wind speeds at vulnerable power poles and upgrading aging infrastructure, would provide additional robustness to protect the grid against extreme winds, especially for extreme winds of long duration (Li and Kopp 2023). For the transition to a weather-driven power system, more consistent and comprehensive data collection procedures and models are needed to inform decisions for system hardening.

More research in forecasting wind events and the advantages of statistical analysis or downscaling methods for power system planners and operators should be considered for wind power generation, installation and design, equipment cooling strategies, and infrastructure planning for both on and off-shore wind.

Multiple strategies exist to predict, prevent, and detect wildfires associated with power systems reviewed in Jazebi et al (2020). Predictive models hold significant merit in preventing and predicting wildfires. At the same time, advances in satellite imaging and sensing, stationary, and mobile ground-based solutions are advancing wildfire detection (Guikema et al 2006).

Limitations to previous studies include using non- comprehensive vulnerability thresholds for power system security. This includes metrics such as N-1 security, meaning the network can maintain normal operations if a single contingency failure occurs. Research should consider implementing PSPS on realistic networks and investigate ways to speed up the optimization solution (Rhodes and Roald 2022).

Another adaptation measure is under-ground power lines in wildfire-prone areas, explored in Taylor and Roald (2022). An optimization problem selects which power lines should be prioritized for under-grounding in a region with significant wildfire risk and cost considerations. Despite wildfire risk variability across different days, the method consistently chooses high-risk lines for under-grounding. For physical infrastructure, table 2 provides guidance for practical system hardening recommendations to protect the system from or during wildfire events.

System hardening is the predominant adaptation measure for hurricanes. These measures include those from 2 referring to extreme winds, flooding, and modernization. Hurricanes pose unique complications, as vulnerable substations may need to be relocated, elevated, or placed in waterproofed rooms, which could change the layout of the grid.

Smart grid technologies are suggested to aid in power recovery and response efforts for hurricane events. As such, two research areas regarding hurricanes include demand-side management and adaptive topology control. Customer participation in DR, which is not typically associated with hurricanes, could aid in servicing critical loads, such as emergency buildings or hospitals. Adaptive topology control could aid power flow rerouting for critical infrastructure by sophisticated switching for transmission and distribution components. Powering critical infrastructure is especially important during hurricanes, as in the case of Puerto Rico; if a system is destroyed, outages can be prolonged.

Proactive strategies have been tested in Mahzarnia et al (2020) that aid in recognizing critical infrastructure prior to expected hurricane conditions. These strategies should be further explored as they prevent infrastructure from being damaged or highlight damaged infrastructure for quicker response times.

Adaption measures for a weather-driven power system are crucial to the security of the power grid, critical infrastructure, and safety. As shown in figures 3 and 4, the financial burden and personal burden of power system destruction and power outages are substantial and increasing with intensity and frequency, with variances given the region. A summary of the adaptation recommendations for the future, which includes system hardening techniques that are currently being implemented from table 2, and knowledge gaps in the current literature are provided in table 3. While each of the extreme events investigated in this review have unique characteristics, thus unique adaptation recommendations, it is evident that system hardening and weatherization are preliminary actions to consider in expanding and contributing to a robust grid in the face of climate change.