3.1. RQ1: What Does Literature (Peer Reviewed or Otherwise) Discuss Regarding BESS as a Solution to Transmission Congestion?
Several papers examined energy storage and wind integration. When wind power produces too much power for the transmission grid to handle, generation can be curtailed. Nguyen et al. used an AC-power optimal power flow (OPF) model to determine the ideal siting of energy storage assets [11
]*. They found that the addition of energy storage reduces locational margin prices (LMPs). They further found that energy storage assets on a nodal transmission system are optimally-sited at nodes with high LMPs and that siting energy storage at the same node as the wind power was one of the least ideal placements for energy storage. The mere presence of energy storage, whether ideally sited or poorly sited, reduced wind curtailment, but optimal siting also provided TCR, thereby lowering LMPs.
Arabali et al. modeled stochastic wind generation with a probabilistic density function. From there, they built a DC probabilistic OPF to account for the uncertainties of wind generation [6
]*. They calculated the size and location of energy storage using particle swarm optimization in order to minimize two costs, the cost of operation and the cost of transmission congestion. Their results show that when wind penetration is low, transmission congestion does not become a major problem. Therefore energy storage is best placed at the same location as the wind generation, which is ideal for preventing wind curtailment. When wind penetration reaches levels high enough to require curtailment even with energy storage, then energy storage is best placed at a bus other than where wind power is generated. This optimal siting allows energy storage to both relieve transmission congestion and cut operating costs by reducing curtailment. Giving the optimization model the ability to distribute storage among two nodes allowed for more cost reductions. Arabali et al. used compressed air energy storage (CAES) assumptions in calculating their model.
Vargas et al. examined transmission congestion management considering the interactions of wind curtailment, energy storage, and ramp-up/down rates of wind and conventional generators [12
]*. Vargas et al. considered the operational costs incurred from re-dispatching conventional generators, noting that generators with slower ramp rates incurred higher costs. Their model assumed that energy storage was co-located with wind generation. Vargas et al. created an OPF model to analyze management strategies for transmission congestion including doing nothing, curtailing wind, using energy storage alone, and using energy storage in conjunction with wind curtailment. They further performed a sensitivity analysis to determine the effects of generator ramp rates and energy storage characteristics. They determined that strategies involving wind curtailment with or without energy storage could eliminate transmission line congestion. They also found that transmission line utilization increased when energy storage was used in conjunction with curtailment. Their sensitivity analysis showed that generators with slower ramp rates created large over-costs, and that energy storage systems (ESS) could reduce the re-dispatch of conventional generation.
Alqunun and Crossley used a mixed integer program to compare distributed energy storage against centralized energy storage in a nodal network [13
]*. They used a DC-OPF and combined economic dispatch with unit commitment problems to minimize operation costs of thermal generation and operation and maintenance costs of energy storage. They tested their model on a six-bus system with three gas generators and used lead-acid batteries for energy storage. They performed test cases with no storage, centralized storage at a single bus, and distributed storage at each of six buses. Their model indicated that both storage cases outperformed the no-storage case, and that distributed storage outperformed centralized storage. They also discovered less transmission and distribution congestion when using distributed storage, which could defer the upgrade of transmission and distribution (T&D) infrastructure.
While some articles arbitrarily sited energy storage [12
]*, other articles analyzed optimal siting of storage systems [6
]*. Dvorkin et al. examined energy storage from the outlook of ensuring profitability in the presence of low-cost wind availability [14
]*. They built a model for siting, sizing, and operation of an energy storage system in the presence of wind generation and conventional generation. They used a bi-level program for optimization. Their upper-level program made siting and sizing decisions, while their lower level program made operational decisions based on representative days. Their program calculated its own LMPs, which allowed it to account for the effects of energy storage on LMPs. They then turned this program into a mixed-integer linear programming (MILP) problem, and solved it using computer simulation. They found that coordinated distributed energy storage brought greater profits to the energy storage owner and savings to the system operator. They additionally showed privately-owned energy storage could draw additional profit by manipulating the LMPs to promote arbitrage, but that the overall system would be negatively effected.
Pandzic et al. devised an optimization model for siting, sizing, and operation [15
]*. They built a three-stage unit commitment problem formulated as a MILP. In the first stage, they allowed variable siting, sizing, and operation. In the second stage, they took the best sites from stage one and allowed for variable sizing and operation. In the third stage, they took the best sizing and siting from stage two and optimized for the best operation strategy. They demonstrated an advantageous connection between energy storage and wind penetration, which reduced LMP volatility. They demonstrated that the feasibility of energy storage was reduced when the price of energy storage increased. They showed distributed energy storage had advantages over centralized storage, but they showed limitations to the gains made from distributed storage after a point. Pandzic et al. performed sensitivity analysis on their model, finding that location and sizing did not change significantly with +/−5% changes to wind generation.
Rosso and Eckroad investigated battery energy storage to provide TCR by relieving thermal constraints on transmission lines [2
]*. They analyzed energy storage used as part of a corrective control scheme. They suggested that in the event of a fault, a battery placed at either the sending or receiving end of the line could prevent a contingency from occurring by absorbing or injecting power. Thus a battery could permit greater transmission capacity through a transmission line by protecting the line from going above its thermal limits. Rosso and Eckroad determined optimal siting of a battery through an analytical heirarchical process model, taking into account sensitivity factors of locations that led to congestion. After determining an optimal site for the battery, the battery power was then calculated from the desired boost in transmission capacity. They observed that adding a 50 MW battery to a transmission bus could increase the pre-contingency capacity of that line by roughly 55 MW.
Khani et al. crafted a real-time optimal dispatch (RTOD) algorithm to optimize privately-owned energy storage revenue [16
]*. In their RTOD, energy storage revenue came from providing TCR and participating in energy arbitrage, ignoring other possible revenue sources. They modeled their RTOD as a MILP. To participate in transmission relief, energy storage frequently needed to prepare itself by charging. Yet if the energy storage asset were trading on arbitrage positions, it might not be sufficiently charged to provide congestion relief. Khani et al. devised an adaptive penalty algorithm to disfavor energy storage participating in arbitrage when transmission congestion was forecasted. This penalty factor caused the energy storage to prefer charging for TCR over trading on the energy market. Khani et al. states that they assumed the revenue for providing congestion relief will reward the decision to prepare for providing TCR. They used CAES models in their simulation.
Berrada et al. devised a dispatch model to analyze energy storage participating in frequency regulation and energy arbitrage [17
]*. They built their model to compare the operation of gravity storage against CAES and pumped hydroelectric storage (PHS). They used a linear programming model and took historical energy market data from the New York ISO (NYISO). When simulated, the model chose to participate in frequency regulation almost exclusively because higher revenues were typically offered in that market. The dispatch model did show profitability for PHS and CAES systems, but found gravity storage would not be profitable unless it could offer additional benefits like deferring T&D upgrades or avoiding distribution outages.
Lu and Shahidehpour examined the optimal scheduling of batteries as part of a photovoltaic (PV) system from a security-constrained unit commitment outlook [18
]*. Their model had a thermal unit sub-problem and a PV-battery unit sub-problem. PV-battery generation could substitute for thermal generation when thermal generation prices were high, causing thermal generation to revise its unit commitments. The overall problem was solved using a Lagrangian relaxation method and network flow programming. The simple optimization goal was to replace thermal generation with PV/battery generation. Lu and Shahidehpour used their model to show that PV/battery systems could mitigate congestion and reduce LMPs. They also looked at the effects of grid-connected PV/battery systems to both reduce LMPs and reduce unit commitment of thermal generation plants.
Mohsenian-Rad examined large-scale battery investments as price-makers rather than price-takers [19
]*. These deployments were large enough to influence pricing in energy markets. Mohsenian-Rad devised a coupled nonlinear optimization problem for market-moving energy storage, which he then reformulated into a MILP that be could optimized. He provided a deterministic version of his solution, as well as a stochastic version to account for market uncertainties. He optimized profits for the energy storage system, finding that energy storage made profits with and without transmission congestion. In general, the model showed that congestion benefited energy storage profits, but not always, as demonstrated in a few cases. Mohsenian-Rad showed that energy storage provided a service to the grid by reducing generation costs. Though energy storage is not directly compensated for this service, it is able to generate revenue directly through arbitrage during congested periods. His study determined that battery efficiencies as low as 80% were still capable of generating profit. He also determined that dispersed battery deployments performed better than centralized batteries, and that economic bidding proved advantageous over scheduled operation.
Das et al. developed a high-fidelity dispatch model that optimized for both bulk energy and ancillary services for various energy storage technologies, including CAES, PHS, BESS, and flywheels [8
]*. They crafted a high-fidelity production costing model using a bi-level program that first determined unit commitment using a MILP and then determined economic dispatch using linear programming. Their model incorporated cross-arbitrage, the act of taking energy from one market, like an upregulation in the ancillary services market, and selling it into a different market, like the energy market. By accounting for cross-arbitrage, they found their model predicted a value increase of a 100 MW CAES unit to be 3.2 times that of a model that did not account for cross-arbitrage. Das et al. examined the question of whether BESS should be designed to participate in both energy markets and ancillary service markets or to participate strictly in ancillary service markets. They found batteries designed for ancillary services made greater revenues than those designed to participate in both ancillary services and energy markets.
Hartwig and Kockar modeled strategic bidding and dispatch decisions of energy storage in a nodal market [20
]*. They created a bi-level optimization model, the upper level designed to maximize the difference in buy and sell LMPs and the lower level a DC-OPF problem. This non-linear model was further relaxed into a MILP. Hartwig and Kockar examined how generation assets and energy storage assets can game an energy market for maximum profit at the expense of overall grid welfare. Their model examined only the ability to bid into energy markets, not ancillary service markets, and focused in particular on the market incentives in Great Britain. Hartwig and Kockar demonstrated that the addition of energy storage to a market with strategically-bidding generation assets would still provide a net benefit to the system as a whole. They showed electrical grids operated best when neither generation assets nor energy storage assets were gaming the system. Hartwig and Kockar showed that as line congestion became a factor, strategic energy storage bidding was more damaging to the system welfare. They also suggested that incentivizing energy storage based on its capacity could improve the system welfare, allowing energy storage to remain available when needed most.
Kazempour et al. developed a near-optimal, self-scheduling model to compare the economic feasibility of PHS to sodium–sulfur batteries [21
]*. They considered day-ahead participation in energy markets, regulation markets, and spinning reserve markets. Their model demonstrated sodium–sulfur batteries and PHS both made economic sense, though PHS showed a much clearer promise of profitability.
While sitting, sizing, and operation are inherent to the planning process, some papers took a decided focus toward planning. Hartel et al. crafted a MILP to approximate the German T&D grid and determine the feasibility of battery energy storage projects [22
]*. Their study did not include market simulation nor allow batteries to participate in ancillary service markets. Hartel et al. demonstrated that energy storage for the purpose of relieving congestion was not financially prudent up to the year 2025. Additional renewable energy generation and decreasing future battery prices could make future deployments feasible. They also added that pure economic reasons for energy storage ignored the ability of BESS to provide resiliency and grid security.
Babrowski et al. analyzed energy storage on the German grid as a long-term planning strategy [23
]*. They used PERSEUS-NET-ESS (Package for Emission Reductions Strategies in Energy Use and Suppl–NET–Electricity Storage Systems) software, which is a bottom-up model of the German electricity system. Using this software model, Babrowski et al. estimated the present and predicted future costs of batteries, determining how batteries could be integrated into the grid starting in 2015 and going to 2040. They built a simplified weather model, ignoring catastrophic weather events, and simulated the present and future German grid. Based on their models, Babrowski et al. determined that energy storage does not show significant benefits until about 50% renewable penetration, or until 2030 in Germany. As a matter of policy, they suggested waiting to buy energy storage assets until the prices came down.
Huynh used an EPRI-created software, Energy Storage Valuation Tool 4.0, to compare energy storage technologies [5
]*. He compared PHS against battery energy storage, finding battery energy storage practical for the U.S. City of San Diego. He further explained how energy storage projects can replace new transmission, significantly reducing greenhouse gasses in the process.
Qiu et al. developed a model for co-planning energy storage projects along with transmission projects [24
]*. Their model incorporates the choices to build transmission infrastructure as well as to install modular energy storage projects that could be upscaled in future years. They built a stochastic co-planning model that included operational and capital costs associated with new BESS and new transmission installations. They then simulated their model on a 24-bus test system. Their model demonstrated that co-planning battery energy storage projects with transmission line projects reduces the number of new transmission lines built. They also found that when transmission congestion was an issue, BESS was optimally sited close to load that was at risk of curtailment. When congestion was not an issue, BESS was instead optimally placed near wind generation to prevent wind curtailment.
While most papers focused on relieving transmission congestion by operating batteries on the transmission grid, a few papers suggest placing energy storage closer to the customer. Xu and Singh developed a model predictive control-based operating strategy for minimizing purchased electricity costs using smart grid communication and controls [25
]*. These predictive control models were designed for distribution network-connected energy storage, but could be used to benefit either the distribution network or the transmission network. Xu and Singh used sequential Monte Carlo Simulation to test their model on a 24-bus system. They showed energy storage at the distribution level could improve the reliability of the bulk energy grid. Barsali et al. discussed smart grid applications for energy storage at the distribution level as a way to relieve transmission congestion, participate in load leveling, enhance power quality, and improve electrical efficiency [26
]*. Sidhu et al. considered storage on the distribution side of a substation, while Hartel et al. considered storage at both the transmission and distribution level [7
Other publications have presented literature reviews of energy storage research, applications, and projects, or have discussed energy storage in general terms. Yao et al. examined how energy storage technologies are applied globally and noted that energy storage is being used to relieve transmission congestion, enhance renewables integration, provide ancillary services, and defer transmission upgrades [27
]*. Hamid et al. discussed small-scale models for the physics of batteries, but also discussed utility-scale batteries and their uses, including energy time-shifting, ancillary services, and resource adequacy [28
]*. Several review articles have been published in the past five years that address the use of BESS in transmission systems. Luo et al. provided a review of ESS technologies, sorting them into six categories based on the type of energy stored. The paper briefly mentioned the use of BESS for providing transmission stability and transmission upgrade deferral [29
]. Yang et al. provided a review of battery sizing criteria and methods. The review focused on BESS for renewable energy systems and categorizes BESS based on specific renewable energy systems that complement different BESS technologies. The authors mentioned transmission upgrade deferrals and discussed curtailment as a means for addressing transmission congestion [30
]. Poullikkas reviewed different types of BESS applicable to large-scale utility applications. The review highlighted two BESS chemistries as appropriate for transmission stabilization and regulation, specifically lead-acid and flow batteries. The paper is too dated to consider the recent advancements in lithium technologies that have made lithium BESS suitable for transmission applications [31
]*. Zakeri and Syri reviewed literature pertaining to the analysis of life cycle costs for utility-scale ESS, including BESS. The authors considered ESS for three transmission-scale scenarios: bulk energy storage, T&D support services, and frequency regulation [32
]*. May et al. analyzed several large-scale lead-acid BESS projects and identified lessons learned. The authors demonstrated that lead-acid BESS remain technically and economically viable for utility-scale applications. The review showed that these projects are able to provide transmission line support and dampen oscillations within transmission systems [33
]. The review presented herein contributes to this body of knowledge by focusing on transmission-scale BESS, highlighting the potential for addressing transmission congestion through the deployment of large-scale BESS projects.
RQ1 surveyed existing literature in the context of using BESS as a means for providing transmission congestion relief. A significant body of literature was found that demonstrates the feasibility of using energy storage systems to relieve transmission congestion. The literature also makes clear that BESS can be used to provide a wide variety of utility services. Since transmission congestion occurs rarely, BESS dedicated to providing congestion relief can be redeployed during most of a year to provide other services, which would help recover investment costs. Energy storage technologies, particularly lithium-ion BESS, are developing rapidly and experiencing cost declines. The literature discusses ways to address siting, sizing, operation, and dispatch of energy storage systems.
3.2. RQ2: What Technologies Are in Development and Which Ones Have Recently Been Deployed?
Poullikkas detailed various battery technologies in 2013 [31
]*. At the time of publication, Poullikkas noted that large battery energy storage systems used sodium–sulfur batteries. He noted that sodium–sulfur and lithium-ion batteries both have high energy densities. Poullikkas made a qualitative analysis of lead-acid, lithium-ion, nickel-cadmium, sodium–sulfur, sodium–nickel–chloride, vanadium-redox, and zinc–bromine technologies. Yao et al. wrote a similar article in 2016, describing the current energy storage technologies, with a emphasis on the Chinese power grid [27
]*. They examined lead-acid, lithium-ion, vanadium-redox, zinc–bromine, and sodium–sulfur batteries. Rosso and Eckroads reviewed the research work of the Electric Power Research Institute (EPRI) pertaining to TCR and storage, particularly as it applies to thermal relief on the transmission lines [2
]*. When put side-by-side with other forms of storage technologies, they concluded that lead-acid, nickel-cadmium and sodium–sulfur batteries were very attractive from a design flexibility and economical standpoint.
Benato et al. examined the use of large-scale BESS on the Italian Transmission system [34
]*. They examined an energy-intensive sodium–sulfur battery deployment, and looked at two power-intensive installations in Ciminna and Codrongianos with both lithium-ion and sodium-nickel-chloride batteries. They described an operation mode for power-intensive projects, where all available batteries would coordinate to mitigate transmission congestion. Benato et al. presented the performance and cost savings advantages that make lithium-ion batteries well-suited for power-intensive applications and make sodium–sulfur batteries well-suited for energy-intensive applications. Benato et al. further discussed the capabilities of the battery deployments, mentioning power-intensive installations could provide TCR, voltage regulation, and frequency regulation.
Due to the high penetrations of renewables, transmission congestion has increased for the island power systems of both Sardinia and Sicily, which have relatively weak transmission interconnections to the mainland of Italy. Consequently, these island networks suffer from slow frequency response due to a lack of system inertia, which is exacerbated by renewables. Schiavo and Benini presented these cases and discussed the pilot project efforts of TSO TERNA to decrease transmission congestion and increase frequency stability through the strategic installation of BESS [35
]. The pilot projects demonstrated both primary and secondary frequency regulation, as well as black-start capabilities. The authors also discussed the regulatory framework and planning that facilitated the development of these pilot projects. Corona et al. considered the Sardinia system in more detail, specifically considering modernization efforts for the island’s system that would allow for high penetrations of renewables without compromising reliability [36
Sidhu et al. performed a social cost–benefit analysis on the Smarter Network Storage project, a 6 MW/10 MWh lithium-ion BESS, installed at the Leighton Buzzard Primary Substation in Great Britain [7
]*. They measured social costs and benefits using Net Present Value (NPV) to compare the two. Sidhu et al. considered the stacking of assets, allowing for energy storage to provide multiple services to the grid, including frequency response, energy arbitrage, and carbon abatement. They mentioned reducing the maximum peak load during the year was a major benefit of energy storage. They ignored transmission upgrade deferral and TCR from their analysis, because they suggested that TCR requires network congestion that would be fixed by transmission upgrades, and they suggested that transmission upgrade deferral had no net social cost or benefit due to how the transmission system in Great Britain was operated.
In 2010, EPRI prepared a white paper describing various energy storage applications and technologies [3
]*. It rated many energy storage technologies with regard to their level of technological maturity. Among the mature technologies were lead-acid, valve-regulated lead-acid, and sodium–sulfur batteries as well as PHS. Advanced lead-acid, vanadium redox flow, and lithium-ion batteries were rated as demonstration projects. In 2015, EPRI updated its Energy Storage Handbook for the U.S. DOE. This paper classified the relative technological maturities of BESS technologies as roughly the same, but lithium-ion demonstration projects, in particular, had appeared to double [1
]*. In 2017, EPRI’s Kaun and Minear presented a course, “Energy Storage for the Electric Grid,” in which they describe lithium-ion batteries as a family of chemical technologies, some of which are deployed and some of which are still in demonstration [10
Lead-acid technology is the oldest type of utility-scale battery technology. Lead-acid batteries are characterized by short discharge times, limited life cycles, and low energy densities [32
]*. They also tend to require maintenance, and they have environmental impacts associated with lead [10
]*. Southern California Edison (SCE) commissioned a 10 MW, 40 MWh battery project in Chino in 1988, the largest of its time. This battery project was decommissioned in 1997 [37
]*. Puerto Rico Electric Power Authority (PREPA) commissioned a 20 MW battery system in 1993. Due to greater cycling than predicted, the battery system was forced to decommission prematurely in 1999. PREPA would later commission a new 20 MW lead-acid battery system in 2004, but due to a battery fire, had to decommission the plant in 2006. In 2000, STMicroelectronics commissioned a 10 MW lead-acid battery to support an uninterruptible power supply (UPS) in Phoenix, AZ. May et al. provide a comprehensive presentation of recent lead-acid battery innovations as well as a review of several large-scale lead-acid battery projects, including the Southern California Edison Chino project, which was used to demonstrate BESS transmission line support [33
]. The authors conclude that the technology is still competitive in large-scale stationary applications.
More recently, improvements in lead-acid technology have resulted in advanced lead-acid batteries with life-cycles up to 10 times that of the original lead batteries [31
]*. In 2011, the Hawaiian Electric Company commissioned a 15 MW advanced lead-acid battery system to support a wind farm on Oahu. This project was partly funded by a loan guarantee from the U.S. DOE Office of Electricity Loan Guarantee [37
]*. The Maui Electric Company also commissioned a 10 MW advanced lead-acid battery system to support a wind farm in Oahu. In 2013, Duke Energy commissioned a 36 MW, advanced lead-acid battery system in Goldsmith, TX. These batteries were later replaced by lithium-ion batteries in 2016.
Nickel-cadmium batteries are among the older battery technologies. They provide high energy density and low maintenance with life cycles heavily dependent on depth of discharge [32
]*. This literature review identified just one example of a large-scale nickel-cadmium battery. The Golden Valley Electric Association built a 27 MW nickel-cadmium battery system in Fairbanks, AK, in 2003 [37
]*. In the event of an unplanned generation or transmission power outage, 15 minutes of power at rated nameplate capacity may be discharged into the grid from this battery. This battery system was still in operation as of 2019.
Lithium-ion batteries have high energy and high power densities, with typical AC efficiencies between 80% and 92% [10
]*. However, they do have life cycle limitations and safety concerns. Lithium ion technologies include a number of different chemistries including lithium manganese oxide, lithium cobalt oxide, lithium iron phosphate, lithium titanate, lithium nitric oxide, and lithium aluminate. From 2011 to present, the largest number of 10 MW+ battery projects were lithium-ion batteries. Between 2011 and 2018, the authors’ survey identified 22 installations in the U.S. that were operational [37
Sodium–sulfur batteries were developed in the 1980’s by NGK Insulators [32
]*. They operate at high temperatures, 570–660
]*, and serious damage can occur when temperatures drop below 150
]*. This survey could find no sodium–sulfur batteries deployments over 10 MW operating in the U.S. Multiple large-scale sodium–sulfur deployments were found in both Italy and Japan [37
Flow batteries are a relatively new technology. These batteries store energy in an electrolytic solution, which undergoes redox as it is pumped through the space between the battery anode and cathode [32
]*. Consequently, ratings for power and energy can be designed independently of one another, as energy is determined by the amount of solution stored and power determined by the active area of the cell compartment. These batteries require the additional complexities of electrolyte pumping and storage[10
]*. Flow batteries have a variety of chemistries, including vanadium redox, zinc–bromine, iron-chromium, and polysulfide-bromide [32
]*. This paper found only one large-scale operational flow battery. Hokkaido Electric Power commissioned a 15 MW vanadium redox flow battery in Abira-Chou, Japan in December 2016 [37
]*. In 2015, Kazakhstan’s Samruk Energy announced a partnership with Primus Power (a U.S.-based start-up) that would allow it would install a 25 MW zinc–bromine hybrid flow battery in Astana, Kazakhstan. In 2016, the China National Energy Administration announced it would install a 20 MW vanadium redox flow battery.
It has been observed that high energy ESS often have slow ramp rates, while high power ESS tend to have fast ramp rates. Given that two different ESS technologies may have such complementary characteristics, several researchers have proposed hybrid energy storage systems (HESS). Hemmati and Saboori provided a review of these technologies, wherein they discussed various HESS configurations, control strategies, and power electronics architectures [39
]. They also presented applications, including coupling with renewable energy systems to improve stability and reliability. The authors did not discuss HESS in the context of transmission congestion. However, the applicability is evident. A large-scale energy storage system, such as pumped storage, that can provide peak demand mitigation during long periods of transmission congestion would compliment well with a smaller, fast-acting BESS that can provide ramp rate support and frequency regulation.
RQ2 considered what technologies are in development and which ones have recently been deployed. A list of battery deployments that met the inclusion criteria are listed in Table 2
along with their technologies and year of commissioning. Researchers are investigating several different battery technologies. The most recent deployments found for nickel-cadmium and lead-acid batteries were in 2003 and 2011 respectively. Nickel-cadmium and lead-acid technologies may well be supplanted or relegated to niche markets by lithium-ion technology. Sodium–sulfur batteries, while not deployed in large-scale in the U.S., still have advantageous features particularly for providing energy services rather than power services to the grid. While flow batteries are still a developing technology, large-scale flow batteries have been demonstrated in Japan and Kazakhstan. Less developed battery technologies include metal air batteries and solid-state batteries [10
]*. This SLR did not find significant information on either of these technologies.
Lithium-ion BESS are the most promising technology for providing transmission congestion relief. The U.S., in particular, has seen an expansion of lithium-ion BESS projects at sizes large enough to influence transmission. The successful development of the 22 lithium-ion projects mentioned in this literature review lead the authors to conclude that lithium-ion BESS are commercially-viable solutions for providing utility-scale energy storage services. Lithium-ion technology has become the preferred choice for utility-scale BESS deployments. Rapidly declining capital costs will increase the likelihood that this technology will be the dominate BESS technology for utility-scale applications in the future.
3.3. RQ3: What Are the Projected Installation and Operation Costs?
Sidhu et al. looked at many of the costs associated with BESS, including capital costs, operational costs, and degradation costs [7
]*. Capital costs included the batteries themselves, the balance of plant, engineering, and construction. Operational costs included maintenance, administration, and control systems. Degradation costs included degraded performance due to cycling and to aging of the battery, a 6 MW, 10 MWh lithium-ion unit. They calculated the total cost of the battery in 2013 as $
14.1 million. Then they used a Monte Carlo simulation to estimate that future costs of the battery would drop to between $
12.0 and $
9.88 million by 2017–2020.
Benato et al. examined the costs of different BESS systems [34
]*. They described lithium-ion installations costing roughly $
1500/kW and $
1500/kWh, sodium-nickel-chloride installations costing roughly $
3000/kW and $
1200/kWh, and sodium–sulfur installations costing $
3800/kW and $
540/kWh. Benato et al. further suggested that these prices make lithium-ion batteries well-suited for power-intensive uses while sodium–sulfur batteries were better suited for energy-intensive uses. Palone et al. describe operating strategies for managing transmission congestion using sodium–sulfur batteries, specifically for transmission lines subjected to high penetration of wind generation [38
Cutter et al. estimated costs for battery energy storage in a range from $
4200/kW, while not explicitly accounting for costs per kWh [40
]*. They estimated installed capital costs of $
3345/kW, and fixed operating and maintenance costs at $
30.80/kW, escalating 2% each year. Cutter et al. used a generalized understanding of battery technologies in their calculations, not specifying a specific technology. Rosso and Eckroads estimated advanced lead-acid battery installments cost of $
875/kWh, inverter costs of $
220/kW, balance of plant costs of $
520/kW, and O&M costs at 1.5% of investment costs [2
In the Qiu et al. model for co-planning transmission and battery energy storage investments, battery storage projects are assumed to be $
500/kW plus $
25/kWh, assuming a lifetime of ten years [24
]*. These are general assumptions of battery performance, without specifying a particular technology. They further assume that the price of a new 138 kV transmission line is $
927,000 per mile with a lifetime of 60 years.
Zakeri and Syri compared performance and cost characteristics of BESS and other forms of energy storage with one another based on data found in literature [32
]. Table 3
describes the average comparative costs of some of the technologies they examined.
Different from projected cost, break-even capital cost shows the maximum price at which an energy storage system would still be viable. In a project report for the California Energy Commission, Fioravanti et al. analyzed past regulation market data from CAISO (California Independent System Operator) [42
]*. They determine the break-even cost of a 20 MW, 5 MWh BESS participating in the CAISO market over a 20 year period is $
17.6 M. Fioravanti et al. state the average cost of 2-h lithium-ion batteries is $
3500/kW and $
1750/kWh. They also state the average cost of 4-h lead-acid batteries is $
3900/kW and $
In an NREL white paper from 2013, Denholm et al. determined that a typical break-even costs for an 8 hr, 300-MW energy storage system providing load-leveling only was around $
300/kW and around $
1400/kW for a system providing load leveling and capacity. They also determined the break-even costs of an 8 hr 100 MW energy storage system to be around $
500/kW for providing spinning reserves only, around $
900/kW for providing frequency regulation only, around $
1700/kW for providing both spinning reserves and capacity, and around $
2300/kW for providing both frequency regulation and capacity [43
RQ3 addressed the projected installation and operation costs of BESS. BESS are priced by power (MW) and energy (MWh), and these costs vary widely by technology. To lesser degrees, replacement costs, and fixed and variable O&M costs also affect total system costs, with wide variations between system types. Several sources have noted the declining costs for BESS, and project additional decreases in the coming decade. Lithium-ion systems have been identified as particularly likely to experience continued decreases in costs, and thereby a very suitable technology for a wide variety of utility-scale energy storage applications. Several sources noted the break-even costs for BESS, and showed the break-even cost point increases as a BESS is dedicated to provide two or more services. As these technologies continue to mature, costs are likely to continue declining. Since the viability of BESS projects are critically dependant upon accurate cost estimates, continued research into current and projected costs is needed so that developers have reliable means for projecting project costs.
3.4. RQ4: What Are the Projected Revenue Streams or Savings Associated with BESS?
Many use cases have been identified for energy storage systems that can be monetized to produce revenue or incur savings for system owners. The DOE/EPRI Electricity Storage Handbook in Collaboration with NRECA identified use cases for energy storage, listed in Table 4
]*. Additionally, Barsali et al. described additional uses for BESS including active filtering, islanding support, reduction of short term interruptions, and reduction of flicker [26
]*. These individual use cases could provide some sort of revenue stream or savings, and some of these revenue streams could be stacked [3
]*. Many of these revenue streams were elaborated further in the literature.
Taylor explained energy storage could be used to buy and sell energy through arbitrage, buying at one point in time and selling back that energy at another point [44
]*. In this case, energy storage profits were easy to calculate as energy storage generated revenue through the temporal differences in nodal prices. Dvorkin et al. showed that the ideal place to put energy storage is at buses with the largest differences in LMP throughout the day [14
]*. This placement promoted energy storage participation in energy arbitrage, making revenue by purchasing energy at a low price and then selling that energy at a higher price. Energy markets exist to facilitate the buying and selling of electricity. Energy arbitrage is typically exercised in either a day-ahead (DA) market or a real-time (RT) market. The DA market approximates the expected energy consumption and production for the next day. Generators make generation offers, load aggregators make demand bids, and based on these transactions, LMPs are determined. The RT market, or “spot” market, corrects the deficiencies in the predictions of the DA market. LMPs in the RT market are determined by the immediate energy needs of the system. The changes in the hourly LMPs of energy in these DA and RT markets offer the arbitrage positions for energy storage to profit [17
]*. Additionally, energy storage can often influence the LMPs, by making strategic offers to keep LMPs favorable for arbitrage albeit at the expense of the electrical grid as a whole [20
]*. This strategic bidding is known as gaming [44
]*, and the cost of this gaming on the social welfare of the system as a whole is known as the price of anarchy [20
A strict look at peak rates and off-peak rates would not fully explain the arbitrage positions of batteries. The efficiency of the storage system must also be considered. An EPRI white paper described this loss: “With efficiency losses of 20%, the battery buys an additional 20% energy during the off-peak period” [3
]*. The processes of storing and releasing energy come at the price of efficiency losses during each transaction. Round-trip efficiency is typically specified for storage technologies, especially for systems with frequent use [2
]*. Round-trip efficiency declines over time in relation to charge-discharge cycles and depth of discharge, and it is a crucial parameter for BESS value calculations [1
]*. A list of round-trip efficiencies used in the EPRI White Paper are noted in Table 5
While energy storage could recuperate costs for providing TCR through energy arbitrage, energy storage would need to charge in advance of providing TCR. Energy storage that was designed to earn money only through energy arbitrage would frequently fail to charge in advance of transmission congestion. Khani et al. noticed this problem with the energy storage dispatch algorithm and created an adaptive penalty to cause energy storage to favor charging to relieve forecasted transmission congestion rather than optimize for energy arbitrage opportunities, as previously noted in Section 3.1
]*. They further mentioned that for this adaptive penalty algorithm to be implemented in practice, system operators must provide better compensation for TCR than the energy storage asset could earn through arbitrage.
Taylor described a method for valuing passive storage in which energy storage did not buy and sell like a generator, but rather was utilized as part of the overall grid system, like transmission lines [44
]*. From a philosophical level, he likened transmission and storage, suggesting that transmission moves electricity spatially while storage moves electricity temporally, that is to say forward in time. Taylor elaborated on this similarity, suggesting that energy storage usage could be compensated passively through storage rights just as transmission was compensated through flowgate transmission rights. While transmission rights were limited to power capacity rights (as in how much capacity the transmission system had available), energy storage could be viewed as having both power capacity rights and energy capacity rights, the former referring to how much power the energy storage could discharge and the latter referring to how much energy the system could store. Taylor modeled power markets linked by energy storage by solving a multiperiod, linearized OPF problem. He then demonstrated how financial storage rights could be used to monetize energy storage in place of arbitrage. He described “storage congestion” occurring when the energy storage asset is charged to capacity. He then demonstrated that financial storage rights could reduce gaming interactions between generators and loads to improve social welfare. Taylor admitted that financial storage rights only described load shifting, and thereby ignored potential ancillary services for which the batteries could be used.
Energy storage can also bid its capacity into ancillary service markets such as frequency regulation or frequency reserve markets. Typically, ancillary service markets purchase the use of energy storage assets by way of a reservation payment for availability and a performance payment when the asset is called into use [45
]*. The most valuable of the ancillary service markets tends to be the frequency regulation market [17
]*. A frequency regulation service earns revenue for reservation capacity and for performance when the energy storage asset is dispatched. A frequency regulation service could receive 400 dispatch calls per day. A frequency reserve service, on the other hand, earns revenue for reservation capacity, but dispatch calls come roughly just 20 times per year. Revenue streams for regulation and reserve service depend on the rules of the transmission operation [45
RTOs and ISOs have devised mechanisms for addressing these various versions of frequency services. PJM, an RTO in the Eastern U.S., devised performance-based frequency regulation, which splits the frequency regulation market in two parts, a RegA market and a RegD market, using a signal filtering process. The RegA market provides a low-pass regulation signal for traditional generation assets that are limited by their ramp rate. The RegD market provides a high-pass regulation signal more appropriate for assets that have high ramp rates, such as BESS. The California ISO, CAISO, distinguishes between up-regulation and down-regulation bids. CAISO offers a reservation payment, a performance payment for utilization, and an accuracy payment for how close the regulation service matches the prescribed set point.
Energy storage can provide revenue indirectly through savings. For instance, when an energy storage asset defers investment in a new wires transmission project, a cost savings may occur. Rosso and Eckroads analyzed energy storage projects from a savings standpoint [2
]*. By relieving transmission congestion and thereby allowing for power from cheaper generation sources, the batteries saved money over base generation. They also considered the capital cost comparison of reconductoring transmission lines to installing new battery systems. DNV-GL performed an analysis of energy storage options as part of Pacific Power’s 2017 submittal [46
]* to the Oregon Public Utilities Commission. DNV-GL, in its analysis of Pacific Power’s grid, calculated the cash flows associated with implementing an energy storage system to both provide frequency control and offer investment deferral. Navigant, in its analysis of Portland General Electric’s (PGE) vertically-integrated grid, calculated the NPV associated with placing energy storage projects at various levels of PGE’s electrical grid [47
]*. A summary of the savings per kWh is listed in Table 6
. Of particular interest to transmission was Navigant’s analysis of a 20 MW energy storage system connected directly to the transmission grid. They tested a case where the energy storage provided transmission deferral 10 days out of the year and resource adequacy another 10 days out of the year while participating in energy and ancillary service markets for the other 345 days. The energy storage system, in this case, showed partial revenues and savings from deferring additional transmission and generation spending, selling into frequency regulation and other ancillary services markets, and participating in energy markets.
System upgrade cost deferral is a major benefit of energy storage. Berrada et al. cite the value of energy storage to defer transmission and distribution upgrades can range from $
]*. When comparing energy storage against conventional generation, energy storage could offer fuel savings. Denholm et al. compare a system with no energy storage to one with energy storage, and demonstrated that energy storage reduced total fuel costs and total start costs (for starting a generator) [43
]*. Greenhouse gas emissions are often considered in energy storage decision-making internationally [2
]*. In California, Huynh figured greenhouse gas emissions to install a 50 kW, 4 hr battery system to be 152 tons of carbon dioxide per GWh of energy [5
]. Huynh compares this to a natural gas peaker plant, producing 469 tons of carbon dioxide per GWh of energy.
Sidhu et al. examined social costs and benefits [7
]*. In this sense, the net social benefits are the revenues of social cost/benefit analysis. These include frequency response, arbitrage, network support, reduced distribution curtailments, carbon reduction, and the terminal value of the battery plant after decommissioning. By considering these values together, Sidhu et al. measured the total social benefits case for installing energy storage. Sidhu et al. ignored transmission deferral in their analysis because they viewed it as a transfer payment.
For RQ4, the projected revenue streams and savings associated with BESS were investigated. These values are difficult to measure due to the many ways BESS services can be combined and the situations under which they are deployed. As such, estimates of value range widely. Additionally, a BESS may be designated to provide a primary service, but its value may be increased by providing additional services when not being dispatched to provide its primary service. This stacking of services, however, is difficult to quantify, due to the stochastic nature of dispatching these services, and particularly if transparent markets do not exist that can express value for these services. Research into mechanisms for cost recovery should continue. As new means for dispatching batteries for grid support services are devised, these additional revenue streams will serve to economically justify utility-scale BESS deployment.
3.5. RQ5: What Are the Economic Indicators Used for Evaluating BESS?
Several economic cost metrics have been adopted or developed specifically for energy storage systems, including net present value (NPV), internal rate of return (IRR), levelized cost of electricity (LCOE), levelized cost of capacity (LCOC), levelized cost of storage (LCOS), cost of new entry (CONE), and flexible cost of new entry (FCONE). This review presents literature pertaining to each of these metrics and discusses their relative values.
LCOE has traditionally been applied to generation assets as a means for providing cost comparisons between multiple capital investment options. An LCOE analysis ascribes all future costs of a system to the present value, which provides a present price-per-unit energy value in $
/MWh. Akhil et al. defined LCOE as “the $
/MWh revenue for delivered energy needed to cover the life-cycle fixed and variable costs and provide the target rate of return based on financing assumptions and ownership types” [1
]*. Obi et al. developed a method for calculating LCOE for various energy storage systems, specifically for the purpose of weighing energy storage systems against one another [50
]*. Their analysis includes sensitivities of LCOE to technical factors including storage efficiency and system lifetime as well as economic factors including discount rate and capital costs. Obi et al. did not consider the ability of batteries to sell ancillary services to the grid, nor did they factor in the energy storage depth of cycling, which can be a considerable factor for BESS as it degrades over the lifetime of the asset.
Hartel et al. examined storage options from the outlook of using storage to prevent curtailment [22
]*. To compare the various storage methods, the paper used LCOE, though without going into any detail into how they calculated it. Zakeri and Syri used LCOE to compare energy storage options against each other [32
]*. They presented cost comparisons between storage technologies based on considerations like discharge time, use (whether bulk energy sales or transmission and distribution support), and discount rates. They considered the use of batteries to include bulk energy, T&D support, and frequency regulation. Zakeri and Syri determined that PHS and CAES plants outperformed battery technologies. However, the project sizes were extremely large and the data they used were current as of 2014.
Schmidt et al. demonstrated the utility of LCOS, which values the discounted cost of discharged electricity for specific energy services [51
]. In contrast, LCOE values the discounted cost of one type of service generated electricity. Schmidt et al. applied LCOS to nine different storage technologies and considered the use of these technologies for providing twelve different energy services over a time period between 2015 and 2050. Services considered include peak demand mitigation, congestion relief, and several types of frequency response, among others. They conclude that lithium-ion BESS will prove to be the most cost-effective technology for nearly all energy services by 2030 due to rapidly-decreasing capital cost.
Telaretti et al. analyzed the economic viability of transmission-scale BESS within the Italian transmission system, including lithium-ion, lead-acid, and sodium–sulphur chemistries [52
]. They use net NPV and the IRR to estimate project value and provide a comparison between these various BESS technologies. The authors apply parametric analysis to investigate the effects of variations in both electricity prices and peak demand charges. They conclude that for BESS become economically viable, there need to be high peak demand prices as well as large differences between high and low energy prices.
Afanasyeva et al. used an LCOE calculation to compare hybrid solar and gas power plants to one another [48
]*. They examined a plant that used a photovoltaic array with a battery energy storage system to provide primary power but also included a gas turbine to keep up with peak demand. They compared the hybrid plant against standard coal-fired and natural gas power plants, and demonstrated that hybrid plants would outperform the gas and coal plants by 2030. Afanasyeva et al. used a Moroccan location as its case study, and they accounted for carbon emission costs.
Akhil et al. defined LCOC as “the $
/kW-yr. revenue per kW of discharge capacity needed to cover all life-cycle fixed and variable costs and provide the target rate of return based on financing assumptions and ownership types” [1
]*. Akhil et al. further say that LCOC is primarily used for comparing capacity resources like combustion turbines, whereas LCOE is primarily used for comparing energy resources like baseload fossil fuel generation or renewables. While Kaun and Minear admit that LCOE is an appropriate indicator for comparing renewable generation with storage to other generation options, they suggest that LCOE is a confusing economic indicator for energy storage in general because storage does not produce energy [10
]*. Kaun and Minear suggest using LCOC or lifetime project NPV instead. Akhil et al. presented LCOE and LCOC for various energy storage technologies including pumped hydro, CAES, sodium–sulfur batteries, sodium-nickel-chloride batteries, vanadium redox flow batteries, iron-chromium flow batteries, zinc–bromine hybrid flow batteries, and lead-acid batteries [1
]*. In an EPRI white paper, energy storage systems are compared to both combined cycle gas turbines and standard combustion turbines using LCOE and LCOC respectively [3
PGE and Pacific Power both submitted cost analyses to OPUC as part of Oregon’s storage mandate [53
]*. When Navigant, on behalf of PGE, modeled a 20 MW energy storage system connected to the transmission grid, it used NPV to compare the value of these projects against others [47
]*. When providing the service of resource adequacy and transmission asset deferral, the energy storage system had an NPV of $
1058/kW for 2 h of energy capacity and $
1649/kW for 4 h of energy capacity. DNV-GL, on behalf of Pacific Power, also used NPV for its cost comparisons of energy storage participation, particularly investment deferral and frequency regulation [46
Cutter et al. used net market revenues, cost of new entry (CONE), and their novel indicator, flexible cost of new entry (FCONE) to compare the market value of energy storage options against a traditional combustion turbine. They described an improvement to the CONE economic indicator that accounts for the ramping flexibility of energy storage equipment; FCONE considers ramping speed of the generation asset [40
]*. They compared BESS, CAES, and PHS against a traditional generation asset, the simple gas combustion turbine. Cutter et al. compared the assets by simulating them as price-takers. They used a MILP to optimize dispatch for each asset bid, applying historical CAISO market data. Examined from a strategic bidding standpoint, the energy storage equipment bid into the day-ahead energy and ancillary services markets while the combustion turbine primarily bid into the more volatile real-time market. From a net revenues perspective, the energy storage assets had almost complete market participation, bidding mostly into the ancillary services market. Combustion turbines tended to have a high bidding price that precluded them from day-ahead markets. Using traditional CONE, the combustion turbines outperformed BESS. However, when ramping rates were taken into consideration, BESS outperformed combustion turbines, CAES, and Pumped Hydro for ramp rates of one second or less.
For RQ5, the economic indicators used for evaluating BESS were considered. Several quantifiable metrics were presented that allow comparison between multiple storage projectsas well as against traditional generation technologies. These metrics include LCOE, LCOC, LCOS, NPV, and CONE, which are commonly used. Several researchers argue that LCOC is a better cost metric than LCOE since the latter is designed for evaluating the value of traditional energy-producing generation assets rather than storage systems. LCOS has been shown to be an effective metric for evaluating various energy storage systems based on the ancillary services that storage systems can provide. FCONE, a modification of CONE, was proposed as a means for accounting for the value of power ramp rate, a characteristic that is particularly advantageous for BESS, which have very high ramp rates. The economic viability of utility-scale BESS projects must be quantifiable as well as comparable to alternative, non-storage projects. The proliferation of economic indicators is enhancing the quantification of BESS projects. However, these different metrics are not directly comparable to each other, and each have their merits and deficiencies. A standardized set of metrics would allow for on par analysis between different ESS options as well as against non-storage options.
3.6. RQ6: Are Storage Models Available That Can Model Added Value of BESS, Such as Arbitrage, TCR, and Ancillary Services?
Several evaluation tools have been developed that can model the economic value that BESS can provide. These tools can account for multiple BESS value streams, but no one package is comprehensive enough to cover all potential cases. Many evaluation tools have been developed by researchers, and a few are available as commercial products.
In an NREL technical report, Denholm et al. described a modeling tool for simulating energy storage at the transmission level [43
]*. They modeled transmission zonally, and accounted for energy storage along with other mixed assets included coal, combined cycle gas turbines, and standard gas turbines. Their model accounted for ancillary services and energy market sales.
Fioravanti et al. developed models for testing energy storage use cases in a project report for the California Energy Commission [42
]*. They used PLEXOS, a software package that simulates unit commitments, to estimate the pricing in the CAISO market. In one case, they modeled transmission energy storage against frequency regulation. In another, they model transmission energy storage against traditional base and peaker plant dispatch.
Hyunh used EPRI’s Energy Storage Valuation Tool (ESVT) 4.0 to analyze the benefit of energy storage to the U.S. City for San Diego [5
]*. EPRI created ESVT as a simulation tool to provide a cost–benefit analysis of energy storage projects. Hyunh considered San Diego acting as a community choice aggregator to provide customers with cheaper power while still relying on San Diego Gas and Electric for T&D services. Huynh determined that the use of small batteries either at the customer level, the substation level, or the transmission level can provide a net savings to power customers while also deferring the need for new transmission lines.
Kaun and Minear used an EPRI created software package, StorageVET [10
]*. StorageVET models grid services including TCR, T&D investment deferral, arbitrage, capacity, arbitrage, regulation, reserves, voltage support, black start, reactive power control, time of use charge reduction, and demand charge reductions [55
]*. While it was originally built for analyzing energy storage in California, it can be reconfigured for analysis in other regions. StorageVET is capable of forecasting prices using historical data and price curves. StorageVET essentially performs cost/benefit analysis on a single use case or a stacking of use cases. StorageVET is a work-in-progress product, so some of StorageVET’s models are not fully developed.
In response to Oregon HB 2193 [54
]*, OPUC issued an order requiring its major utilities, Pacific Power and PGE, to analyze the value of uses for energy storage [53
]*. To perform this analysis, PGE commissioned engineering firm Navigant. Navigant analyzed PGE’s balancing area using the Navigant Valuation of Energy Storage Tool (NVEST). Navigant performed five test cases, connecting energy storage at various levels within PGE’s network: within transmission, at a distribution substation, on a distribution feeder, and to large customers (like commercial/industrial) and small customers (like residences). Each storage use case was modeled individually using NVEST. Individually, capacity was the most valuable use for energy storage while TCR was deemed to have almost no value, as PGE does not frequently suffer from this problem. Navigant used an analysis that stacked energy storage values, but Navigant eliminated transmission congestion, distribution congestion, distribution upgrade deferral, black start, voltage support, and reactive power control from its analysis because “these applications were considered to have low value” [47
Berrada et al. developed an optimization method to test the feasibility of gravity storage compared to PHS and CAES, though the method could be applicable to battery storage [17
]*. Their model allows energy storage to participate in regulation service or energy markets, either day-ahead or real-time. The model showed a clear advantage to participating in the regulation control market, rather than buying and selling in the market. They ignore reserve service, as they assumed regulation service would pay higher revenues.
Kazempour et al. developed a near-optimal scheduling method to compare the economic feasibility of PHS to sodium–sulfur batteries. Their model considered energy markets, regulation markets, and spinning reserve markets and bidding into these markets on a day ahead basis [21
RQ6 investigated the available BESS models that account for added-value services such as economic arbitrage, TCR, and several types of ancillary services. Available modeling tools can capture some, but not all of the potential revenue streams from energy storage, and some allow for analysis of stacked services, wherein a BESS can be committed to providing multiple services concurrently. EPRI’s software StorageVET may be the most complete of these software packages, capable of modeling multiple service cases, though many of its models are not yet fully developed. The utility industry would benefit significantly from open-source modeling software that could account for the intricacies of large-scale BESS. Further, modeling software should provide APIs so that BESS models can be integrated with utility simulation packages such as GridLab-D, CYME, PSSE, and PowerWorld simulator.