A Cost-Optimizing Analysis of Energy Storage Technologies and Transmission Lines for Decarbonizing the UK Power System by 2035

Abstract

The UK net zero strategy aims to fully decarbonize the power system by 2035, anticipating a 40–60% increase in demand due to the growing electrification of the transport and heating sectors over the next thirteen years. This paper provides a detailed technical and economic analysis of the role of energy storage technologies and transmission lines in balancing the power system amidst large shares of intermittent renewable energy generation. The analysis is conducted using the cost-optimizing energy system modelling framework REMix, developed by the German Aerospace Center (DLR). The obtained results of multiple optimization scenarios indicate that achieving the lowest system cost, with a 73% share of electricity generated by renewable energy sources, is feasible only if planning rules in England and Wales are flexible enough to allow the construction of 53 GW of onshore wind capacity. This flexibility would enable the UK to become a net electricity exporter, assuming an electricity trading market with neighbouring countries. Depending on the scenario, 2.4–11.8 TWh of energy storage supplies an average of 11% of the electricity feed-in, with underground hydrogen storage representing more than 80% of that total capacity. In terms of storage converter capacity, the optimal mix ranges from 32 to 34 GW of lithium-ion batteries, 13 to 22 GW of adiabatic compressed air energy storage, 4 to 24 GW of underground hydrogen storage, and 6 GW of pumped hydro. Decarbonizing the UK power system by 2035 is estimated to cost $37–56 billion USD, with energy storage accounting for 38% of the total system cost. Transmission lines supply 10–17% of the total electricity feed-in, demonstrating that, when coupled with energy storage, it is possible to reduce the installed capacity of conventional power plants by increasing the utilization of remote renewable generation assets and avoiding curtailment during peak generation times.

1. Introduction

The UK government has been at the forefront of climate action, committing to become carbon neutral by 2050. In response to this pledge, the Net Zero Strategy was launched in 2021, consisting of a ten-point plan to reduce emissions across each sector of the economy over the next three decades. One aim of this strategy is to fully decarbonize the UK power system by 2035, subject to security of supply, with a forecasted 40–60% increase in electricity demand, reaching a load of 500 TWh/year. This is based on 50% electrification of the UK’s car fleet and the installation of heat pumps in 13 million homes by 2035 [1]. The strategy includes key milestones to achieve this goal, such as installing 40 GW of offshore wind by 2030, producing 10–17 GW of green hydrogen by 2035 for fuel supply, and deploying flexibility options including energy storage. The Future Energy Scenarios report published by the National Grid estimates that 15–36 GW of energy storage will be required by 2035 to support the stabilization of the power system [2]. Although this figure provides a general idea of the overall energy storage requirements, it does not offer an in-depth analysis of the technologies best suited for the task or the associated implementation costs. Understanding and analyzing the role of different energy storage technologies remains an area of potential study with a view to realizing grid decarbonization commitments.

Y. Scholz, H. Kondziella, M.R.M. Cruz et al. agree that the main challenge facing future power system grids is the need for significantly greater flexibility than what exists today. These grids must be capable of reacting quickly and reliably to overcome the high intermittency associated with electricity generation from renewable energy sources, while also satisfying highly variable electricity demand, thereby achieving a balanced and stable power system [3,4,5,6]. Traditionally, such flexibility has been supplied by dispatchable spare power plants that can be connected to the grid when required. However, this “spinning reserve” is usually expensive and polluting. Integrating Energy Storage Technologies (ESTs) with renewable energy generating assets helps overcome the challenge of intermittency by offering a much less carbon-intensive alternative for balancing the power system. EST provides instant frequency control, with response times ranging from seconds to milliseconds, as well as bulk storage for daily and seasonal dispatch. In many countries, electricity consumption is only one part of the overall energy demand, with heating/cooling and transport holding significant shares. Therefore, linking these three sectors by transforming excess electricity produced from renewable sources at times of peak generation into other energy carriers is key to avoiding curtailment and maximizing the returns on investment in renewable electricity generation assets.

Power to X (P2X) addresses this by converting power into other forms of energy, including the following:

• P2Mobility: Using Electric Vehicle (EV) batteries as floating daily energy storage that can be returned to the grid during times of low generation [7].
• P2Gas: Transforming electricity into renewable hydrogen produced from electrolysis, allowing seasonal energy storage when large quantities are stored underground as compressed gas [8].
• P2Heat: Electrifying heat generation with the use of heat pumps and P2Ammonia to produce ammonia as a hydrogen carrier for long distance transport [9,10].

Future power systems should be interconnected, featuring single electricity markets that remove technical and commercial barriers to trade with neighbouring regions and countries. This would allow sharing electricity surpluses through interconnections with regions that require it during times of low renewable electricity generation, avoiding curtailment at the site of excess generation while reducing overall installed capacity [7,8,9,10,11]. Price or incentive-based Demand Response (DR) schemes are also crucial for adding flexibility to the demand profile, enabling consumers to have a more active role in their electricity demand, thereby shaving load peaks and filling valleys to flatten the demand curve [12].

Currently, several ESTs are available to provide grid balance, each with different applications according to power rating and discharge time. High-power supercapacitors, flywheels, and Superconducting Magnetic Energy Storage can provide frequency control due to their quick response, that is, within milliseconds to seconds. Batteries have a slightly slower response, within several seconds to minutes, and capacity ranges that make them more suitable for daily energy storage. On the far-right end of the chart, Pumped Hydro Storage (PHS), Compressed Air Energy Storage (CAES), and the use of hydrogen as an energy carrier are more suitable for seasonal bulk power management [13].

It is important to note that while some technologies, such as PHS and CAES, have been used for many years and are well known, several others are in development or demonstration phases. For example, Adiabatic CAES is an enhanced version of traditional CAES that captures the heat generated when compressing air underground with a heat exchanger, to be utilized later during decompression at the time of discharge, thereby increasing the overall process round-trip efficiency. Other technologies, such as flywheels, sodium–sulphur batteries, and lithium-ion batteries, are currently in the deployment phase, while flow batteries, supercapacitors, and superconducting magnetic energy storage are in the demonstration stage. Furthermore, hydrogen underground storage coupled with fuel cells is in the development stage, with some facilities being built around the world to test performance and scalability. Technology maturity is an important factor because less mature technologies introduce uncertainties in terms of performance, feasibility at commercial scale, and system costs.

It is expected that future energy systems will operate under completely different regimes compared to current ones, mainly due to the uncertain and intermittent nature of electricity generation from Renewable Energy Systems (RESs). These systems introduce high fluctuations in various timeframes (i.e., seconds, daily, and seasonal), as opposed to the traditional dispatchable nature of fossil fuel-based power plants that offer more stable and predictable profiles [14]. With increased shares of RESs, significant changes in electricity pricing patterns are anticipated, driven mostly by fluctuating generation profiles. This will result in low prices during peak generation hours and high prices during low output hours/seasons (i.e., nighttime and off-season periods). This new dynamic will make energy storage systems more financially attractive to developers and investors, as electricity fed to the grid at those times can be sold at premium rates [15]. Anastasovski et al. analyze the role of electricity, heat, and chemical energy storage in developing positive energy districts, discussing implementation challenges including high upfront investment costs, a lack of effective incentive strategies for deployment, and the need for smart systems to optimize grid integration [16].

From the demand side, load is expected to be as dynamic and variable as generation, due to the high electrification rates of sectors that currently rely heavily on fossil fuels, such as transport, heating, and cooling. This will inevitably increase overall demand and alter typical daily and seasonal profiles [17,18]. Climate change itself is expected to contribute to seasonal variability, with anticipated hotter summers and colder winters. On the other hand, EVs will play a dual role in the smart cities of the future, acting as electricity demand when charging for driving purposes, but also becoming portable energy storage that can be discharged to the grid during peak load times. The implementation of Demand Response schemes, coupled with bi-directional charging infrastructure and smart metering, are crucial elements to realize the potential of Vehicle to Grid [19].

Due to the anticipated demand variability and the need for grid balancing across various timeframes, energy models must have at least an hourly resolution, or less if available, to adequately incorporate the effects of flexibility options to balance the power system in response to intermittent generation [7,20,21,22]. Another important aspect to consider when modelling decarbonized power systems is that electricity generated from renewable resources must have priority in the merit order to avoid curtailment, thereby maximizing asset returns [15,20]. Additionally, the use of dispatchable ESTs for charging and discharging, followed by electricity feed-in from interconnections, and, lastly, relying on fossil-fueled “peaker” plants coupled with Carbon Capture and Storage (CCS) as the last available supply resource should be considered.

Given the target set by the UK government to decarbonize the power system by 2035 and the new dynamics expected to be faced by these modern power systems that feature high shares of intermittent renewable energy generation, this paper aims at building an energy model that simulates these dynamics to answer the following two critical questions: firstly, how does the integration of energy storage and grid interconnections help reduce the dependence on conventional power generation for the UK power system? Secondly, is it possible to determine the generation, storage, and interconnection power capacity mix that implies the lowest cost for the UK power system?

Modelling energy systems has attracted a lot of attention over the last two decades, with a view to understanding how various flexibility options can support balancing the power grid when there are high shares of renewable generation. Most studies performed so far fall mainly into one of the following three categories:

• Conceptual studies that analyze the applications and limitations of various flexibility options or examine the features and capabilities of currently available models to simulate and understand how future energy systems would function.
• Implementation of energy models to microgrids.
• National or regional energy system modelling.

Due to the nature of this paper, the focus was on the first and last categories. J. Bistline [14], J. Després et al. [23], and A. Fattahi et al. [7] analyze and compare 19 Integrated Energy System Models, discussing the key elements these models need to address to accurately represent future national energy systems. These key features include hourly or higher temporal resolution; the integration of EST to balance high intermittency; sectoral coupling across electricity, heating, cooling, and transport sectors to allow P2X approaches; cross-border electricity trade by simulating interconnectors; and linkage to macroeconomic models for adequate demand forecasting. This analysis helped with the selection of the most appropriate modelling package for the set objectives [7,14,20,23].

G. Schweiger et al. [10], J. Sijm et al. [11], and H. C. Gils et al. [21] are some of the authors who analyzed, through energy modelling, the behaviour of future decarbonized energy systems, considering intermittent generation profiles for electricity, heating, and P2Gas systems at a national or regional level, including Sweden, Germany, the Netherlands, the whole European region, China, and Antigua and Barbuda in the Caribbean. Analysis and results vary from one study to the next; however, there is common agreement on the important role of energy storage, open trading markets, and DR mechanisms to minimize curtailment and optimize system performance [10,11,21,24,25,26,27,28].

The literature review concluded that, although few studies have been published to analyze the decarbonization of national and regional energy systems through modelling, some of them evaluate optimal installed capacities for generation, energy storage, and interconnections for the assessed region, and the proposed conclusions are site-specific and only applicable to the assessed geographies given their locally available renewable sources. There is a lack of publicly available energy models analyzing the optimal pathway to decarbonize the future UK power system by 2035. This gap motivated the development of this paper, which is aimed at discussing the technical and economic impact of integrating energy storage systems and interconnections into the power system when there are large shares of intermittent renewable generation, with a focus on answering the two critical questions proposed.

The rest of this paper is organized as follows: Section 2 covers the Materials and Methods, describing the high-level architecture of the energy modelling framework, key input parameters, and model assumptions. Section 3 presents the model results, including a sensitivity analysis. Section 4 includes the Discussion and critical analysis of the obtained results. Lastly, Section 5 recaps the Conclusions and key findings.

2. Materials and Methods

The analysis of this work is based on the Renewable Energy Mix (REMix) energy modelling framework, developed by the energy systems modelling group at the German Aerospace Center “Deutsches Zentrum für Luft und Raumfahrt” (DLR). This framework was chosen because of its open source code; hourly and regional resolution that is ideal for simulating national power systems with high shares of intermittent renewable generation; its flexible modular approach, allowing for the definition of multiple technologies for generation and energy storage [29]; its cost-minimization objective function for optimizing outputs; and, lastly, its proven reliability in generating comparable results to other national energy systems models [24]. REMix runs within the General Algebraic Modelling System (GAMS) environment [30], which consists of a mathematical programming language that integrates with commercial solvers such as the IBM CPLEX for optimization [31]. The model is formulated with a linear, deterministic cost-minimization objective function, along with a series of constraints, to ensure the power system remains balanced every hour of the simulated year, generating the following outputs: system commodity balances, optimum capacities by technology type in each region, and associated system cost, given the defined model inputs [3,21,25].

Figure 1 shows the high-level framework block diagram, providing a quick overview of the main inputs, the optimization objectives, and the outputs generated. The system was modelled in three stages, building up complexity by adding flexibility options with each stage:

• Stage 1: Represents the basic power system composed of the estimated 2035 load and the available power generation technologies for each of the UK regions. In this stage, each region behaves as an independent grid.
• Stage 2: Builds upon stage 1, integrating energy storage technologies.
• Stage 3: Builds upon stage 2, adding grid interconnections between the UK regions and with neighbour countries such as Ireland, Belgium, France, and the Netherlands.

Model Inputs

The model inputs are described in detail in Figure 2. Four UK regions were defined, including England (ENG), Wales (WL), Scotland (SCT), and Northern Ireland (NI), each one modelled as an independent node. Four additional nodes representing relevant neighbour countries were introduced in stage 3, when interconnections are integrated to the system, including France (FR), Belgium (BE), Netherlands (NL), and Ireland (IE).

For each of these nodes, hourly load profiles were derived from the estimated 2035 electricity demand, considering key parameters such as population growth and forecasted electrification of transport and heating sectors. On the supply side, normalized hourly generation profiles corresponding to intermittent renewable generation, according to natural resource availability for each of the defined regions, were considered. Renewable energy technologies include solar Photo-Voltaic (PV), run-of-river hydro (Hydro), and onshore and offshore wind. Dispatchable generation technologies were also available for each node, including Combined-Cycle Gas Turbines (CCGTs) coupled with Carbon Capture and Storage (CCS) to minimize emissions, as well as nuclear and biomass. Seven energy storage technologies were also defined according to their appropriate application, including Supercapacitors (SCaps), Fly Wheel Systems (FWSs) for frequency control, lithium-ion batteries (LIBs), and redox vanadium Flow Batteries (FLBs) for daily load shifting; Pumped-Hydro Storage (PHS), Adiabatic Compressed Air Energy Storage (ACAES), and underground hydrogen storage (H₂) produced with Reversible Solid Oxide Fuel Cells (RSOFCs) suitable for long-term seasonal energy storage. Minimum and maximum capacity ranges were set for each technology in each node, and aligned with realistic estimates published for the defined regions, allowing the solver to optimize the capacity mix according to the set cost-minimisation objective function.

Provided that the Net Zero strategy aims to decarbonise the power system by 2035, all scenarios and estimations refer to that year as a target. Electricity demand distributions by region were estimated based on three components, as listed below, each one following the corresponding hourly profile proposed by the Energy Plan 2020 UK country model [32]. As a result, it was forecasted that by 2035, approximately 509 TWh of electricity will be demanded by the four UK regions, including the following demands:

• Domestic and industrial electricity consumption, forecasted from population growth trends [33,34,35,36] and estimated annual load per capita for the four regions [37,38]: Based on an analysis of data from 2009 to 2019, a decreasing trend in electricity consumption per capita in the range of −1.7% to −2% was identified for all UK regions. This trend was assumed to continue until 2035 assuming further implementation of energy efficiency measures;
• Heating electrification corresponding to 30% of households [1];
• Transport electrification corresponding to 50% of the car fleet [1] and 13% of the bus fleet.

A total of seven generation technologies are mapped for optimization; the fuel conversion parameters considered for the three dispatchable technologies are listed in

Table A1. Fuel conversion parameters.

Technology	Fuel Type	Fuel Efficiency [%]	Input Fuel [GWh]	Fuel Cost [MMUSD/GWh]	Fuel Limit [TWh]
CCGT with CCS [64]	NG	47 [64]	2.13	0.0382 [65]	No limit
Nuclear [66]	U235	35 [67]	2.85	0.0046 [68]	No limit
Biomass [64]	Wood	29 [64]	3.45	0.0308 [69]	170 [70]

. The remaining four generation technologies refer to intermittent renewable generation, including solar PV, hydro, and onshore and offshore wind. The normalized hourly generation profiles were estimated per region from Typical Meteorological Year (TMY) timeseries, obtained from public world atlas databases. PV and hydro normalized profiles are directly proportional to the renewable energy source availability. In the case of PV, the profile is associated with global horizontal irradiation [39], and for hydro, it is derived from total precipitation [40]. Normalized onshore and offshore wind generation profiles were estimated from source data for wind speeds at a 10 m height (U_R) [40]. Wind power conversion efficiency, given in Equation (1), was resolved for power output by building the synthetic performance curves shown in Figure A1 and Figure A2 based on selected onshore and offshore turbines suitable for UK winds, as shown in

Table A2. Onshore and offshore wind turbine parameters.

Wind Data	Turbine Diameter [m]	Hub Height (Z) [m]	Rated Power [MW]	Cut-in Speed [m/s]	Cut-out Speed [m/s]	Air Density [kg/m3]
Onshore	150 [71,72]	165	6 [71,72]	3 [72]	25 [72]	1.225
Offshore	236 [73,74]	260	14.7 [73,74]	3 [74]	30 [74]

. Wind shear, given in Equation (2), was used to extrapolate the wind speed timeseries from a 10 m height to the corresponding hub heights; according to the reference turbines chosen for onshore and offshore wind farms, typical atmospheric stability was assumed (m = 0.14) [41].

Wind power efficiency and shear equations are given in (1) and (2), respectively.

$$\eta ={\displaystyle \frac{P_{out}}{\frac{1}{2}\rho A{U_Z}^3}}$$

(1)

$$U_Z=U_{R }{\left({\displaystyle \frac{Z}{Z_R}}\right)}^m$$

(2)

where the following hold: η: Efficiency; ρ: air density [kg/m³]; A: rotor swept area [m²]; P_out: power output [W]; U_z: wind speed at hub height [m/s]; U_R: wind speed at reference height [m/s]; Z: hub height [m]; Z_R: reference height [m]; m: atmospheric stability constant.

Minimum and maximum power capacity constraints in units of giga-watts (GWs) were defined in every region for each interconnection line and generation and energy storage technology defined, setting the minimum limits equal to the total current operational plus the capacity reported as approved [42,43,44]. In the case of CCGTs, the minimum was set to zero for all regions to let REMix test potential optimums without fossil fuel electricity generation. The maximum capacity constraints were set by assessing the potential deployment by technology for every region of the model, considering limiting factors such as laws that deter or prohibit the deployment of certain energy projects; such is the case of Ireland, where nuclear fission is not allowed for the generation of electricity [45], or in England, where planning rules become an obstacle for the permit approval of onshore wind projects [46]. Economic input parameters are also required for each technology mapped, including operating lifetime, project amortization time, Capital Expenditure (CAPEX), Operational Expenditure per year (OPEX), and project financing rates. Values used for the baseline scenario can be found in

Table A3. Cost parameters for defined conversion technologies.

Technology	Operating Lifetime [Yrs]	Amortization Time [Yrs]	CAPEX [MUSD/GW]	OPEX [MUSD/GW/Yr]	Financing Rate %
CCGTs with CCS [64]	25	25	1715	37.1	7.3
Nuclear [66]	60	30	3600	77	7
PV [64]	35	25	458	11.7	5
Wind_Onsh [64]	25	25	1336	39.5	5.2
Wind_Off [64]	30	30	1611	123.2	6.3
Hydro [66]	80	40	4269	63	5.4
Biomass [64]	25	25	4086	125.7	7.9

. The economic analysis is reported in Million USD (MUSD) for easier international benchmarking; data sources published in different currencies were converted using average exchange rates from 2020 to 2022 [47].

Stage 2 of the model integrated seven energy storage technologies to address various grid balancing needs. SCaps and FWSs were considered for improved power quality, providing frequency control with fast charge/discharge cycles and response times in the range of milliseconds to seconds. LIBs and FLBs were included mainly for daily load shifting, with response times in the range of a few seconds and discharging for several hours during the day. PHS, ACAES, and H₂ were added to provide bulk power management for seasonal storage and increased energy security. The existing halite basins in the UK subsurface are suitable for building salt caverns to store large amounts of energy in the form of compressed hydrogen or air. Underground seasonal energy storage has been a proven method at Teeside in Yorkshire since 1972, where up to 1 million m³ of high purity hydrogen has been stored at 50 bar in three salt caverns to produce ammonia and methanol [48]. These halite formations were considered to estimate the maximum storage capacities of ACAES and H₂ [49,50,51].

Table A4. Energy storage conversion parameters by technology.

Storage Type	Application	Response Time	Lifetime [Yrs]	Round-Trip Eff. [%]	Power/Unit	Energy Storage/Unit
SCap [75]	Frequency Control	0.016 s	16	92	3 MW [76]	17.2 KWh [76]
FWS [75]		0.25 s	20	86	250 KW [77]	25 KWh [77]
LIB [78]	Daily Load Shifting	1–4 s	10	88	100 MW [79]	400 MWh ** [79]
FLB [78]		1–4 s	15	70	300 MW [75]	1.2 GWh [75]
PHS [78]	Seasonal Storage	0.5–8 min	40	80	400 MW [75]	5.5 GWh [75]
ACAES [78]		3–10 min	25	68 [80]	300 MW [75]	5 GWh [75]
H2-RSOFC * [77,81,82]		0.5–10s	30 [83]	40 [82]	35 MW [84]	50 GWh [84]

* Reversible Solid Oxide Fuel Cell (RSOFC); ** utility scale battery bank equivalent to 1 unit.

and

Table A5. Energy storage costs by technology.

Storage Type	Yr of Estimate	Converter CAPEX [MUSD/GW]	Storage CAPEX [MUSD/GWh]	Fixed OPEX [MUSD/GW/yr]	Variable OPEX [CAPEX %]	Financing Rate [%]
SCap [75]	2025	835	66,640	1	1%	10%
FWS [75]	2014	376	3686	4.9 [77]	1%	10%
LIB [78]	2030	71	252	0.71 [26]	1%	7%
FLB [78]	2030	116	307	9.26	1%	10%
PHS [78]	2035	4531	0.011	35.9	1%	5.4%
ACAES [78]	2030	1089	80 [26]	9.82	1%	10%
H2-RSOFC * [77,81,82]	2035	1750	0.21 [48]	61.25	3.5%	10% [83]

* Reversible Solid Oxide Fuel Cell.

include the key parameters in terms of technical performance, cost and capacity constraints for power conversion, and energy storage.

Nine grid interconnections are introduced in stage 3, increasing system flexibility, allowing regions to export excess generation and therefore avoiding curtailment, and enabling electricity imports when required. Power transmission among regions and neighbouring countries is modelled, assuming grid expansions installing HVDC lines, reducing power transmission losses, and modernizing the network to cope with the expected increase in electricity demand. The framework allows one to define whether a transmission line is laid under sea or on land, and the setting of different loss coefficients per kilometre and infrastructure costs. In this stage, four additional nodes were introduced to the system, corresponding to the countries that currently have the largest electricity trading market with the UK, including Ireland (IE), Belgium (BE), France (FR), and the Netherlands (NL) [52]. Three land transmission lines were defined, including ENG-WL, SCT-ENG, and IE-NI. Remaining interconnection lines are mainly laid under sea, but have a section placed on land to connect to substations; these include the following: SCT-WL, SCT-NI, IE-WL, FR-ENG, BE-ENG, and NL-ENG. The 2021–2022 average electricity price of $0.35 MUSD/GWh [53] was assumed to estimate the cost associated with the trading of electricity with regions outside of the UK.

Table A6. Transmission losses and costs [26].

Component	Loss Coefficient (Per GWh of Electricity Flow)	Lifetime [Yrs]	CAPEX	OPEX	Financing Rate
AC/DC station	14.5 MWh/h	40	182 MUSD/station	1.1 MUSD/Line-yr	7%
Land transmission	34 KWh/h km	40	0.43 MUSD/Km	3.5 KUSD/Km-yr	7%
Sea transmission	26 KWh/h-km	40	2.9 MUSD/Km	23.8 KUSD/Km-yr	7%

shows the key parameters corresponding to interconnections modelled.

3. Results

In this section, the results of the baseline scenario are presented in Section 3.1. This includes a comparison of how the electricity is fed into the UK grid by each technology, optimum capacities, and overall system cost change when energy storage and interconnections are integrated into the system in stages 1 through 3. Section 3.2 presents 12 additional scenarios to understand how sensitive the model outputs are to changes in critical input parameters.

3.1. Baseline Scenario

Figure 3 shows the electricity fed into the grid per region and the whole UK by grouped category in exhibit (a) and by technology in exhibit (b). The first observation is that the largest electricity feed-in corresponds to England for all three stages due to having the largest load. It is also interesting to see how total electricity feed-in increases by more than 20%, from 509 to 619 TWh, when energy storage is integrated into the grid, and by more than 44% when interconnections are considered, increasing to a total of 737 TWh.

Electricity feed-in from storage technologies increases from 74 TWh in stage 2 to 79 TWh in stage 3, representing about 11–12% of the total electricity supplied, while in stage 3, 118 TWh is fed in from interconnections, representing 16% of the electricity supply. Another highlight is that dispatchable generation drops from 186 TWh in stage 1 to 71 TWh in stage 3 due to the integrated flexibility technologies.

As observed in exhibit (b), the generation technology that provides the largest share of electricity is offshore wind, located mainly in England. However, when interconnections are added to the system, onshore wind increases mainly in Scotland and Wales, generating an additional 106 TWh that is fed in through interconnections, resulting in a reduction in CCGTs and nuclear generation by a total of 97 TWh.

In terms of the aggregated installed capacity for the four regions, the observed decrease in electricity generation from dispatchable sources in stage 3 is mainly due to a sharp drop in CCGT capacity from 76.3 GW in stage 1 to 7.5 GW in stage 3, as well as a drop in nuclear capacity from 21.5 GW in stage 1 to 8.9 GW in stage 3, and finally a less steep drop in biomass from 18.6 to 11 GW. On the other hand, onshore wind capacity increases from 35.5 GW in stage 1, with 60% of that capacity installed in Scotland, to 41.8 GW in stage 3, by adding 6.3 GW in the Wales region to take advantage of the transmission lines to lower the overall system cost. Offshore wind installed capacity remains unchanged for all three stages at 34 GW, while PV is estimated at 28 GW when interconnections are considered in the system.

The optimization estimates that about 11% of the electricity feed-in is supplied by 3.5 TWh of energy storage capacity. Figure 4 presents the breakdown of storage capacity by technology. As can be observed, underground hydrogen storage corresponds to 88% of the total estimated storage capacity, located mainly in England due to the salt formations located in that region. The second largest storage capacity is PHS, corresponding to 220 GWh, followed by 135 GWh of ACAES and lastly 74 GWh of LIBs.

Regional interconnections play a very important role in decarbonized power systems, helping to maximize the utilization of remote renewable generation assets by transmitting the electricity to the region of demand, thereby reducing reliance on conventional power generation in areas where it is not viable to exploit renewable energy sources. The stage 3 optimization results provide evidence of this statement, estimating that interconnections provide 118 TWh, corresponding to 16% of the total electricity feed-in. As a result, CCGTs and nuclear plants only provide 69 TWh/yr, representing a 9% share of the annual electricity supply.

To better understand the interconnection dynamics in the model, exhibit (a) in Figure 5 plots the electricity imports as upwards blue bars and the exports as downwards green bars for each of the regions. The four UK regions are presented on the left side of the dashed line, and neighbouring countries with shared interconnections on the right side. As can be observed, England is the largest importer, demanding 112 TWh of electricity per year, while Scotland is the largest exporter with 78.2 GWh, followed by Wales and France. Exhibit (b) shows the net electricity flow for each transmission line, indicating in dark green that the majority of the electricity flows from Scotland to England and from Wales to England, while exhibit (c) displays the utilization of transmission lines, indicating in dark red that the interconnections Scotland–England, Wales–England, Scotland–Wales, and Northern Ireland–Scotland have the highest utilization throughout the year.

For the cost-minimization optimization, the following cost categories are considered: capital expenditure (CAPEX), corresponding to the cost associated with building the required infrastructure; operational expenditure (OPEX), related to the operating and maintenance costs for the installed capacity during the defined lifetime; and the fuel costs associated with generating dispatchable power from technologies such as CCGTs, nuclear, and biomass. In addition to these three, UK electricity trading through interconnections with France, Ireland, Belgium, and the Netherlands was also considered to estimate the total system cost for each scenario.

Optimization results show a drop in overall system cost from a total of USD 55 billion (BUSD) in stage 1 to USD 51.7 billion in stage 3 when energy storage and interconnections are integrated into the power system. In total, USD 19.5 billion is spent on energy storage, representing 38% of the system cost, of which USD 11 billion corresponds to LIBs. The second largest expenditure, representing 31% of system costs, corresponds to renewable energy generation, with USD 14 billion spent on wind generation capacity. Additionally, 6%, corresponding to USD 3.3 billion, is spent purchasing a net balance of 9.3 TWh of electricity, mainly from France and Belgium.

3.2. Sensitivity Analysis

The baseline scenario (BL) results presented in Section 3.1 correspond to the simulation outputs given the set of input parameters presented in Appendix A. However, it is acknowledged that the values assumed for the baseline scenario rely on currently available sources of information and may vary based on technology adoption, political decisions, or simply overlooked information. Therefore, this section aims at analyzing the model sensitivity to changes in the values of selected parameters, such as the reduction in installed capacity costs for certain technologies that are in early stages of deployment or not yet mature, changes in allowable capacities, lower project financing rates, increased prices of natural gas, and variations in the estimated electricity demand.

Table 1. Scenarios considered for sensitivity analysis.

Scenario	Type	Description
1	Capacity constraint	Increased maximum allowable onshore wind capacity in England and Wales due to relaxation of planning rules.
2a	Generation Cost reduction	Reduced CAPEX and OPEX by 15% for following dispatchable generation technologies: CCGTs with CCUS, and biomass.
2b		Reduced CAPEX and OPEX by 15% for offshore wind.
3a	Energy storage Cost reduction	Reduced CAPEX and OPEX by 30% for supercapacitors and flywheels.
3b		Reduced CAPEX and OPEX by 30% for FLBs and ACAES.
3c		Reduced CAPEX and OPEX by 20% for H2.
3d		+/− 20% CAPEX and OPEX variation for LIBs.
4	Lower finance rates	Homogenization of financing rate to 5% for all generation and storage technologies.
5	Capacity constraint	Increased maximum allowable capacity of interconnections.
6	Cost reduction—interconnections	Reduce interconnection cost by 20%.
7	Fuel price increase	Natural gas price increase by 100%.
8	Load variation	+/− 5% variation in electricity demand.

describes the 12 scenarios that were run to analyze how the optimization results change with the modification of a single set of parameters. The first scenario refers to the relaxation of the current planning rules set in 2016 for the permit approval of onshore wind farms in England [54,55]. The current law requires that the Local or Neighbourhood Plan identifies the site as suitable for wind energy and, following the consultation with local communities, they support the developer’s proposed measures to address the concerns raised by affected residents [46]. The implementation of this law has resulted in a significant slowdown of new onshore wind projects in the last 7 years [56]. This scenario explores the effects of increasing the maximum allowable capacity for onshore wind in England due to the potential update in planning rules to achieve decarbonization commitments. Scenario 7 is another interesting scenario that looks at the impact of doubling the price of natural gas, as experienced in the early months of 2022, resulting from the restrictions imposed by the European Union on Russia due to the armed attacks on Ukrainian territory.

Given that the optimization framework is set to optimize the output based on a cost-minimization objective function, the sensitivity analysis presented in Figure 6 shows the difference in total system cost between each of the scenarios defined and the baseline scenario. As can be observed, the most cost-effective power systems correspond to the ones proposed by scenarios 1, 5, and 8.

For scenario 1, the optimum total onshore wind capacity is increased from 42 to 53 GW, making electricity imports unnecessary due to self-sufficiency and allowing the export of a total of 35.5 TWh, thereby helping offset the overall system costs by nearly USD 15 billion. In the case of scenario 5, the optimum total interconnection capacity is increased from 32.8 GW to 49.2 GW, allowing the UK to become a net electricity exporter, selling 19.7 TWh mainly to Belgium and the Netherlands. Lastly, when the demand is reduced by 5% in all the UK regions, the country can once again sell a net 2.7 TWh of electricity, thereby reducing the system costs by USD 5.6 billion.

It is important to note that these are the only three scenarios that result in the UK being a net electricity exporter. In all other scenarios, the UK is instead a net electricity importer, ranging from 2.2 TWh in scenario 2a to 21 TWh in scenario 7 when the price of natural gas is doubled.

When analyzing the electricity feed-in for all scenarios, as shown in Figure 7, it can be observed that generation from dispatchable power plants reaches the lowest share in scenarios 1 and 5, remaining below 7% of the total supply. Renewable electricity and interconnections are very stable for almost all scenarios, averaging 63% for RES and 16% for interconnections, except for scenario 1, when the RES supply is maximized, reaching a 73% share, and minimized for interconnections, representing only 10%. Storage capacity remains stable for all scenarios, averaging 11% of the total supply.

Energy storage capacity is quite variable throughout the scenarios, as shown in Figure 8, ranging from 2.4 to 11.8 TWh. It is the lowest in scenario 2a when the cost of CCGTs and biomass are reduced by 15%, and the highest in scenario 3c when the cost of hydrogen is reduced by 20%. Clearly, the model relies heavily on underground hydrogen storage for all scenarios, allocating to it 82–97% of the total storage, corresponding to 2.1–11.5 TWh, followed by PHS, ranging from 2 to 9%, though remaining stable at 220 GWh for all scenarios. The third largest storage technology is ACAES, ranging from 0.6 to 6%, it being the lowest in scenario 3c and the highest in scenario 3b when the cost of FLBs and ACAES is reduced by 30%. It is interesting that for scenarios 3c, 4, and 7, when energy storage capacity peaks, it is possible to completely eliminate the supply from CCGTs.

The charts presented in Figure 9 provide a detailed cost analysis considering all scenarios assessed, showing that the UK power system cost ranges from $36.7 to 56.2 BUSD. On average, 38% of the total cost is spent on energy storage, 31% on renewable generation, 22% on dispatchable generation, 2% on transmission, and 7% on electricity purchases from neighbouring countries.

Energy storage costs range from USD 17.2 to 21.8 billion, of which 59% corresponds to LIBs, 19% to ACAES, 12% to H₂, and 10% to PHS. Transmission line infrastructure costs range from USD 0.9 to 1.4 billion, while electricity trading with nearby countries can be much more dynamic from one scenario to another, either spending another USD 0.7–7.4 billion to purchase electricity or generating an income of USD 0.9–12.5 billion by becoming a net electricity exporter, as discussed for scenarios 1, 5, and 8a.

4. Discussion

This section aims to analyze the obtained results in the broader context, assessing their reliability and relevance by discussing how they compare to the actual UK power system and how feasible it would be to achieve the proposed capacity gaps. Uncertainties and model limitations are also discussed to evaluate how they impact the obtained results.

4.1. Results Analysis

The findings presented in Section 3 show a good match with major aspects of the actual UK power system. As evidenced in Figure 3, England is allocated the largest supply of electricity in all modelled stages to meet the corresponding demand, which is proportional to the largest population in the UK (84% of people live in England [33,34,35,36] and 84% of the domestic and industrial electricity is demanded by this region [38]). In stages 1 and 2, when regions behave as independent grids, the electricity feed-in share for England is 84%. However, in stage 3, when interconnections are considered, it drops to 69%. This drop is caused by the ability to import electricity from other regions.

Figure 5 shows that England’s imports come mainly from excess electricity generated by Scotland’s 21 GW onshore wind generation capacity, which remains unchanged for all three stages. In stage 3, Wales increases its onshore wind capacity from 2.2 to 8.5 GW, exporting most of its excess electricity to England as well. To put things in perspective regarding electricity transfers, Scotland exported to England an average of 16 TWh/yr between 2019 and 2022, with a power transmission capacity of 3.2 GW [57,58]. However, the model estimates that 60 TWh of electricity needs to be transferred by 2035 with an optimum transmission capacity of 8 GW by this interconnection alone, which is more than three times the current average.

Currently, the UK is a net importer of electricity generated in France, Belgium, and the Netherlands. This continues to be the case in the 2035 model, except for the Netherlands, which would import 1.4 TWh from England. France would sell 10.6 TWh to England, and Belgium 1.8 TWh.

Figure 10 helps in understanding the gap between the confirmed installed capacity (operational capacity plus planned projects) and the scenario results. The red arrows highlight the minimum capacity increase required for each technology indicating that the main gaps relate to building at least another 12 GW of onshore wind, mainly in England and Wales, an additional 14 GW of LIBs, 13–22 GW of ACAES, 7–9 GW of H₂ as estimated for most scenarios, and at least another 16 GW of interconnections between the regions. It can also be observed that by increasing the mentioned installed capacity, it is possible to eliminate at least 32 GW of CCGT and 12 GW of nuclear energy.

According to the model results, hydrogen underground storage with fuel cell reconversion plays a very important role in the future UK energy system, despite its low round-trip efficiency and high CAPEX and OPEX converter costs. This is explained by the high energy storage capacity offered by underground salt reservoirs, making the overall cost per unit of energy stored low compared to other alternatives. This can be evidenced in Figure 9, where it is shown that 12% of the total storage expenditure corresponds to H₂ despite representing more than 80% of system energy storage capacity and feeding to the grid 8.5–34 TWh/yr depending on the scenario, which is very much possible with the estimates of 2200 TWh storage capacity by J. Scafidi et al. [59].

The above-mentioned findings also correlate well with the conclusions of a study performed to analyze the energy storage requirements in Europe by 2050, where it was concluded that hydrogen would be the dominant energy carrier in the UK, with 11–16 GW of converter capacity required, along with high shares of wind generation and major grid expansions between the UK and France in the range of 5–10 GW [26]. Two of the milestones of the UK Net Zero strategy are to achieve 40 GW of offshore wind by 2030 and 10–17 GW of green hydrogen production by 2035, which compare quite well with the results. The total estimated storage converter capacity ranges from 61 to 77 GW, which is on average 2.7 times higher when compared with the storage capacity estimates in the future energy scenarios published by the National Grid.

As mentioned in the sensitivity analysis section and shown in Figure 10, the model relies heavily on generation from onshore wind due to its cost-effectiveness, justified by the electricity yield given the technology costs. However, project developers are preferring to invest in offshore wind farms instead, due to easier and faster permitting given the current planning rules in England. Another important factor is the scalability and accelerated development of larger, more efficient offshore wind turbines to maximize yield, with relatively marginal additional investment.

Taking this information into consideration, scenario 2b was run, and interestingly, even reducing the cost of offshore wind by up to 20%, REMix does not increase the installed capacity of this technology. Therefore, this demonstrates that offshore wind generation is not the most economical way to decarbonize the power system given the current price and energy conversion capacity.

4.2. Uncertainties and Limitations

It is important to highlight that the 2035 estimated demand does not consider the additional electricity that would be needed to produce the green hydrogen potentially required for applications other than power generation, such as transport or heating. This demand was not included in the estimate because there is a high degree of uncertainty regarding the timeframe and scale at which the hydrogen economy might start and gain traction, given the current lack of infrastructure to support the transport and industrial sectors. Therefore, if the hydrogen economy manages to kick off successfully and grow within the next few years to support these sectors, the forecasted electricity profiles would be underestimated.

Solid oxide electrolysis cells (SOEC) have been in development for a number of years, enhancing electrochemical performance and overall stack durability over time. SOECs have proven to have better conversion efficiencies than other technologies, lower material costs, and the capability to handle dynamic load profiles, which is a critical feature for P2Gas applications. This technology has advanced significantly, and in April 2024, Topsoe A/S announced plans to build a 1 GW SOEC for the production of renewable hydrogen in Chesterfield, Virginia, US, after the allocation of funds by the federal government under the Inflation Reduction Act Law [60,61]. However, RSOFC technology is behind compared to solid oxide electrolysis and fuel cells, with a Technology Readiness Level (TRL) of 4–5, and focus is on improving thermal management, enhancing electrode materials that are less sensitive to impurities, and optimizing the reversing time to reach steady-state conditions when switching from electrolysis to fuel cell mode and vice versa [62,63].

Given the scenario results with renewable generation contributing up to 63% of the total electricity fed into the grid, converted mainly from the highly intermittent wind energy resource, it is strongly believed that the role of frequency control technologies such as flywheels and supercapacitors, which have response times in the range of milliseconds to seconds, is greatly underestimated due to the hourly resolution of the modelling framework. This means that the model is simply unable to identify the requirement for balancing technologies with almost instantaneous response. To do so, the time resolution of the framework would have to be much higher, in the range of minutes, including the input data such as regional load and generation profiles.

The current analysis assumes a flat electricity tariff for electricity trading with other countries, which is an oversimplification of the actual trading schemes, given that tariffs are dynamic based on spot electricity supply and demand. To make the model a little more realistic in this aspect, it would be interesting to implement a dynamic tariff scheme and analyze the impact on optimization results.

As discussed in the introduction, it is expected that future power systems will also be flexible from the demand side, with consumers having a greater power of choice to defer electricity consumption to another time of the day or the week, driven by demand response policies and price incentives. REMix does not yet consider this aspect to have a more dynamic management of the demand profile, which can be considered a current model limitation.

4.3. Project Contribution

The main contribution of this paper is that it provides a detailed analysis of the optimum capacity mix that allows the decarbonization of the UK power system by 2035 in line with the Net Zero strategy. It also provides insights into the dynamics of integrating high shares of intermittent renewable generation, with multiple energy storage technologies and interconnections between regions and neighbouring countries, to achieve power balance and guarantee energy security. Government entities in the UK in charge of energy infrastructure planning, policymakers, and energy infrastructure developers may benefit from the findings of this energy modelling.

There is an opportunity to expand the scope of this study by incorporating dynamic electricity tariffs, demand response measures, and bi-directional controlled charging of the Electric Vehicle fleet to add more flexibility options to the system and analyze their impacts on optimum installed capacity.

5. Conclusions

At the beginning of this paper, two questions were formulated to summarize the paper’s objectives. The first question aimed to discover if the integration of energy storage and grid interconnections would help reduce the current dependence on conventional power generation. As discussed in Section 3.1, adding these flexibility options to the power system can increase the electricity feed-in from renewable sources up to 73%, supplying up to 535 TWh/yr. This is supported by 71–91 TWh discharged by various energy storage technologies, mainly underground hydrogen storage, which helps capture large amounts of excess generation from wind power, and 74–133 TWh of electricity shared between regions/countries through transmission lines. Ultimately, this helps reduce the need for dispatchable electricity generation from 186 TWh to 71 TWh, dropping the required CCGT installed capacity from 76.3 to 7.5 GW and from 21.5 to 9 GW for nuclear generation.

The second question aimed to determine the optimum power system capacity mix to achieve decarbonization of the grid by 2035. As discussed in Section 4, at least another 12 GW will need to be built to reach 42 GW of onshore wind capacity in the UK. Should the planning rules be relaxed in England and Wales to facilitate the development of onshore wind projects, the UK total capacity for this technology may reach 53 GW, allowing the country to become a net electricity exporter. Regarding energy storage, an average of 4.4 TWh is required, with underground hydrogen storage having the largest share. In terms of storage converter capacity, the scenario results indicate that on average, 33 GW of LIBs are required, along with 20 GW of ACAES, 9 GW of H₂, and 33 GW of interconnections.

The sensitivity analysis indicates that relaxing the planning rules for onshore wind projects in England and Wales is necessary to achieve a 73% share of renewable energy generation while also achieving the minimum system cost (compared to all scenarios run) by allowing the UK to export a total of 35.5 TWh of electricity to neighbouring countries. Increasing interconnection installed capacity from 33 to 49 GW or reducing the electricity demand by 5% also allows the UK to become a net electricity exporter, thereby offsetting the overall system cost given the export income. It was also identified that the use of CCGTs with CCS is not part of the energy mix when the hydrogen storage cost is reduced by 20%, or when project financing rates are set to 5% regardless of the underlying technology, indicating that the banking sector has a significant role to play in supporting the energy transition by providing preferential rates when financing energy storage, power transmission, and low-carbon energy projects. In summary, decarbonizing the UK power system by 2035 costs in the range of USD 37–56 billion, of which 38% is energy storage and 31% renewable generation.