TIM: Modelling pathways to meet Ireland’s long-term energy system challenges with the TIMES-Ireland Model (v1.0)

,


Introduction
Ireland faces very significant challenges in meeting greater energy needs in the future with a much lower carbon footprint.
Ireland has a high per-capita carbon footprint relative to the European Union average (e.g. in 2018, 13.5 vs. 8.6 tonnes CO 2 e per capita (European Environmental Agency, 2022)) and was expected to fail its 2020 decarbonisation objective as set by the 2.2 TIMES Model generator TIMES (The Integrated MARKAL-EFOM System) is a bottom-up optimisation model generator for energy-environment systems analysis at various levels of spatial, temporal and sectoral resolutions (Loulou et al., 2016a, b).The TIMES code, written in GAMS and available under an open source licence (IEA-ETSAP, 2020) is developed and maintained by the Energy Technology Systems Analysis Programme (ETSAP)2 , a Technology Collaboration Programme (TCP) of the International Energy Agency (IEA), established in 1976.TIMES models can have single or several regions, and typically are rich in technology detail, used for medium-to long-terms energy system analysis and planning at a regional, national or global scale.
TIMES is a linear optimisation, techno-economic, partial-equilibrium model generator which assumes perfectly competitive markets and perfect foresight.Model variants enable myopic foresight, general equilibrium, stochastic programming and a variety of multi-objective function options.The standard objective function maximises the net total surplus (the sum of producers' and consumers' surpluses) which, in a perfect market with perfect foresight, equates to maximising the net present value (NPV) of the whole energy system, maximising societal welfare.Profits, taxes and subsidies are internal transfers, i.e. occurring within the economy, that do not change the NPV (albeit taxes and subsidies can be included to influence the optimisation).It calculates the energy system specification which minimises discounted total energy system costs over the model time horizon, which is the sum of investments, fixed and variable costs, fuel import costs and export revenues for all the modelled processes, less potential salvage values of investments for which whole lifetime goes beyond the model time horizon.
The user inputs the following to the model generator: -Reference Energy System (RES), is the process-flow architecture of economic sectors and energy flows (commodity) between processes (technology), which consume and produce energy, energy service demands and/or other commodities such as environmental emissions (including greenhouse-gases) and other materials.The base-year energy flows are calibrated to national energy balances.
-Energy service demands, are the physical services required by the economy and society for mobility, heat, communications, food etc., which drive energy demand.
-Energy supply curves, the quantities of primary energy resources (e.g.wind power) or imported commodities (e.g.oil, gas, bio-energy) available at specific costs points for differing quality and quantity of energy commodities.
-Techno-economic parameters of existing and potential future energy technologies: economic parameters including current and projected future investment and fixed/variable costs and efficiencies of technologies for energy supply (e.g., solar PV panels, transmission and distribution infrastructure, biorefineries, hydrogen production) and energy demand (e.g., electric vehicles, natural gas boilers, carbon capture and storage); technological parameters including transformation efficiency, availability factor, capacity factor and emissions factor.
-User constraints, which can be any combination of linear constraints (including fixed, maximum or minimum bounds on growth, investment or shares) on technologies or fuels.These are typically used to simulate real-world technology constraints or to simulate policy scenarios.A typical user constraint for decarbonisation analysis is limiting total annual or cumulative CO 2 emissions to model energy system pathways that meet a national decarbonisation target.
TIMES outputs the optimal investment and operation level of all energy technologies which meet future energy service demands at least cost, while respecting user constraints.The model also produces corresponding energy flows, emissions and marginal prices of energy and emissions flows.

Model architecture
Fig. 1 shows a simplified RES in TIM.It describes the structure and energy flows including two major parts: supply-side and demand-side.The former comprises energy resources, fuel production and conversion technologies (e.g.biorefineries, hydrogen production and different power plants), transmission and distribution infrastructure (e.g.gas pipelines and power grid).
The latter covers end-use sectors (e.g.transport, residential) and the corresponding energy service demands (i.e.passenger, freight, hot water).Energy resources incorporate both domestic fossil-based fuels and renewables potentials.These fuels are processed and then distributed across country.End-use technologies consume energy commodities to meet energy service demands.GHG emissions from fossil fuels combustion and process-related emissions in industry are tracked at the fuel supply module, electricity generation technologies and sectoral-consumption levels.The discount rate, the degree to which future values are discounted to the present, is a key parameter in the TIMES objective function.A social discount rate reflects how society views present costs and benefits against future ones and is lower than a financial discount rate, which is how firms make investment decisions.In appraising potential projects or investments, the Government applies a social discount rate.Broadly speaking, in an ESOM scenario with a carbon budget constraint, a higher discount rate would promote later decarbonisation and less capital-intensive technology choices.In this model, a discount rate of 4% is applied, which is based on a Social Rate of Time Preference methodology as set forth in the Public Spending Code (O'Callaghan and Prior, 2018).This rate is consistent with García-Gusano et al. (2016) who recommend using a maximum value of 4-5% for the social discount rate in ESOMs.
Technology-specific discount rates (also known as hurdle rates) are typically used in ESOMs to capture investment decisionmaking from the individual user or industry perspective, to capture market imperfections, limited finance and behavioural heuristics which limit the uptake of novel or capital-intensive investments, even when they are cost-optimal.These parameters are not used in the core version of TIM given its use for modelling long-term energy system pathways from a societal perspective.However, variants of the model can be developed to simulate real-world impacts of policies and behaviour which can include hurdle rates (Aryanpur et al., 2022).

Time & Geography
TIM has been developed with a deep knowledge of the geography of the Irish energy system.A special spatio-temporal approach was taken in the RES base year specification and scenario file data structures to allow flexible regional definitions and temporal resolution in TIM.The model can run in multiple modes with multiple configurations of regional and temporal resolution ranging from a single region national model at a single annual timeslice, all the way to 26 counties at hourly resolution where supply-demand data is available at that spatio-temporal granularity (Electricity, Gas & Transport).
High temporal granularity is needed to appropriately model energy futures with high variable renewable energy systems integration, especially in scenarios with high levels of electrification of end use demands.At the same time, high temporal granularity can be computationally expensive and can significantly increase the time required for model development and testing.In TIM, we address this issue by constraining all timeslice model input data to a single file, generated with a specific temporal resolution.A timeslice tool is used to aggregate raw time series data and created a file in required format (Balyk, 2022).
High spatial granularity is required to give greater policy clarity on optimal investment needs based on region-and countyspecific characteristics to enable counteracting socioeconomic challenges such as energy poverty and infrastructure development within an optimisation framework (Aryanpur et al., 2021).We address this in TIM by creating model input data structures that allow specifying data and formulating constraints on both national and county level.Internal file switches, as well as user shell options, could then be used to apply TIM on either of the levels.
We have decided to focus on this data-driven spatio-temporal model setup to future-proof TIM and enable future Irish energy policy research needs as data-availability improves during TIM's lifespan.

Demands: Drivers & projections
Energy service demands in end-use sectors are driven by growth in the population and in the economy.The model is set up to allow for alternative scenarios for these drivers resulting in different energy service demand projections in the end-use sectors, e.g. as applied in Gaur et al. (2022a, b).This section describes baseline driver projections with a detailed description of energy service demand projections included in the respective sector sections.

Population
Historical population estimates and future projections are obtained from CSO (2020d).We use the M2F1 scenario since it represents a medium growth in population and is in line with population projections used in other national sources (Yakut and de Bruin, 2020).

Economic growth
Economic growth projections are mostly taken from national sources to ensure internal consistency.GVA and income projections are modelling outcomes from Bergin et al. (2017) used by national sources.The following historical and projected data is used: -Historical Gross Value Added (GVA) for the required NACE categories in the services and industry sectors is obtained from EUROSTAT database.Projections for GVA are from the Ireland Environment, Energy and Economy (I3E) model (Yakut and de Bruin, 2020).
-Gross Domestic Product (GDP), both historical and future projections, is obtained from OECD (2018).
-Income, historical values of total incomes, are taken from CSO (2021).Assumption about income growth in the future are from the National Transport Model (AECOM, 2019).
-Modified Gross National Income (GNI*) is derived from CSO's labour force scenario combined with a forecast for output per person (CSO, 2018).(Pfenninger et al., 2018).

Energy service demands
Web-based dashboards3 have been extensively used in the model development process, both for internal model diagnostics and for external engagement and review.The first TIM scenario results archive has also been published on Zenodo (Daly et al., 2021).
TIM has been co-developed with the LEAP-Ireland model (Mac Uidhir et al., 2020), which is a bottom-up simulation model of the Irish energy system which simulates the impact of different policy measures and targets on overall GHG emissions in Ireland, with a particularly granular representation of transport and residential heat demand.Underlying data for the relevant sectors are shared between TIM and LEAP using a Data Repository and shared coding convention dictionary to improve the consistency between the models, give more robust analytical insights for policy and to share and exchange expertise between the modelling teams.This also facilitates multi-model approaches to energy systems modelling, which can make use of harmonised hybrid frameworks coupling simulation and optimisation simultaneously (Rogan et al., 2014).
TIM has also been developed to enable soft-linking with the Ireland Environment, Energy and Economy (I3E) macroeconomic model developed at the Economic and Social Research Institute (ESRI) (Yakut and de Bruin, 2020).I3E is a singlecountry, multi-sector (NACE) inter-temporal computable general equilibrium (CGE) model focusing on environmental and energy accounts in Ireland.While COre Structural MOdel for Ireland (COSMO) focuses on the influence of monetary and fiscal policy on economic activity in Ireland, I3E supplements the macroeconomic outlook from COSMO with environmental and energy disaggregation.I3E retrieves economic growth rates and population estimates from COSMO.TIM derives macroeconomic drivers coupled to the output variables of I3E, enabling scenario variants based on alternative monetary, fiscal and macroeconomic futures, as well as rapid energy system outlook updates aligned with the update cycle of the macroeconomic outlooks from the ESRI.

Overview
The supply sector (SUP) in TIM represents the primary and secondary energy commodities and the processes by which those same commodities are imported, exported, domestically produced through mining or capture of renewable energy potentials and transformed or refined for end-use consumption within the energy system both in the base year (2018) and into the future.
The supply sector declares the future available routes for commodity trade for import and export of energy commodities in terms of quantity of energy, and in terms of import capacity through ports, pipelines and interconnectors at any given time in the model horizon.

Energy Balance and Commodity Declarations
Building the supply sector begins with declaring the energy commodities as per the SEAI (2019) energy balance, as reported to the International Energy Agency.Attention is taken to ensure best practice coding conventions are followed for each commodity -coal, oil, gas, first-and second-generation bioenergy (biogas, bioliquids and solid biomass), liquid and gaseous hydrogen, wind, solar, geothermal, wave, tidal, municipal wastes, agriculture wastes, industrial food waste.Active transport time for walking and cycling is also declared.All base year commodities are declared in the supply sector to maintain clear, tidy and transparent data structures within TIM.
The model development team co-created a convention-dictionary workbook within the main repository that utilises automated three-letter coding conventions for technology, commodity and file name creation.This designed and followed code convention enables clean creation of process-sets, commodity-sets, results-sets and data automation routines for pre-processing and post-processing of data inputs, data outputs and results.Ultimately this makes the model architecture more transparent and easy to develop, enabling new users into the community of developers and peer review.Moreover, setting out predefined and intuitive commodity naming convention and a dictionary that is shared across TIM and LEAP-Ireland has multiple benefits for multi-model coupling, diagnostics and results reporting that is discussed.

Emissions Tracking
The environmental emissions from each primary energy commodity are tracked on the basis of energy and processes-based emissions via combustion and activity-based emissions intensity factors, calibrated on an sector-by-sector basis.Within the supply sector, methane, nitrogen dioxide, sulfur dioxide, carbon dioxide, nitrogen oxide, and particulate matter can be tracked in the sector's emissions accounting balance.In the current version of TIM, biomass is assumed to be carbon neutral.However, future versions of the model may include different assumptions on emissions linked to use of biomass.

Import/Export
Primary and secondary energy commodities, both fossil energy and bioenergy, are imported from international markets at international prices.There are no constraints on the import quantity of oil and coal as it is assumed that international markets can supply domestic demand and that there is sufficient on island storage aligned with IEA-OECD energy security protocols.
Imported gas via pipeline and LPG are modelled on an annual basis.Bidirectional electricity interconnectors to the UK are also represented.Greenlink and Celtic Interconnector are assumed operational from 2023 and 2026 respectively.

Fuel prices
Fuel prices for each imported energy commodity are sourced from IEA (2019, 2020) World Energy Outlook.Secondary commodity import prices are index linked to the primary commodities by a price ratio on the basis of the current ratio between primary and secondary commodity prices.For example, imported gasoline is assumed to be 1.65 times the price of crude oil per unit energy.

Domestic Energy Resources
Domestic fossil fuel reserves for production of natural gas from both the Corrib and Kinsale gas fields, and the production of peat, are calibrated with two-step supply curves.These supply curves are constrained in terms of cumulative energy reserves, annual production costs, and annual production output in energy terms to account for typical production profiles from each field.
Renewable energy potentials (hydro, wind, solar, waste, ocean, geothermal, and ambient heat) are declared in the supply sector but are calibrated and constrained in their relevant sectors, such as power generation and transport.

Bioenergy potentials
Bioenergy potentials are calibrated to SEAI (2015) both in terms of sustainable import volume availability and domestic production potentials.Domestic bioenergy potentials such as sawmill residues, post-consumer recycled wood, municipal waste, tallow, recovered vegetable oil, straw, animal waste and industrial food wastes are modelled with three-step supply curves in terms of price and quantities available.Crop-based bioenergy feedstocks are modelled within the agriculture sector.Domestic bioenergy supply potentials and costs are sensitive to scenarios narrative and as such are modelled within scenario files to account for uncertainty.

Refineries
Ireland has only one oil refinery: Whitegate, in County Cork.It is calibrated to import crude oil at international prices and converts crude oil to refined products limited by upper bounds of output shares such that the output has flexible upper shares of 22.9% gasoline, 8.1% kerosene, 40.9% diesel and 34.4% heavy fuel oil.The refinery is constrained to stay at current capacity, has a lifespan of 50 years, and the production costs are differentiated by production (flow) costs for each output fuel.

Electricity Interconnectors
Existing electricity interconnectors to the UK are calibrated such that exports are priced 30% lower than imports.The 2018 export and import quantities of electricity are calibrated as per historical data.The future activity is constrained using upper bounds, given that TIM does not model the UK electricity market, other than through predefined price signal.This constrain can be relaxed, but is used to explore domestic needs for system flexibility, security, system services and storage.Neither reinforcement of existing nor investment in new interconnectors are currently included in the model.

Biorefineries
Future bio-refinery technologies are defined in the supply sector future technology (SubRES) database.The following technologies are defined: 1) ethanol production from wheat, woody biomass and grass; 2) bio-diesel production from OSR, woody biomass, industrial food waste, RVO and tallow; 3) wood pellets production from biomass; and 4) biogas production from grass, woody biomass, municipal waste, industrial food waste and animal wastes.All bio-refining technologies are defined in terms of start year, efficiency, investment costs, availability factor, and the operation and maintenance costs.

Hydrogen
Hydrogen production is modelled, distinguishing future investments in centralised and decentralised electrolysis options.Delivery options are disaggregated and costed at high pressure for both transmission and distribution pipelines as well as a road tanker option for distribution.Hydrogen storage is also modelled within TIM at the finest timeslice level allowing hourly production and consumption of hydrogen to be represented.This is particularly useful for modelling electricity grid balancing with unit commitment, dispatch and capacity expansion while capturing variable renewable energy system dynamics.

Overview
Ireland has a high share of variable renewable electricity for a relatively isolated grid, with 32.5% of electricity generation in 2019 coming from onshore wind energy.Achieving the 2020 electricity from renewable energy sources (RES-E) target has encouraged strong growth in onshore wind, while increasing the non-synchronous penetration of renewables to 70% by 2030 including offshore wind development is a key policy objective over the next decade as Ireland moves towards a net zero carbon electricity system.

Existing Dispatchable Grid
The existing 61 generation unit fleet is calibrated to the base year of 2018 from the Commission for Energy Regulation (CRU) I-SEM validated PLEXOS model (Geffert et al., 2018).Each existing generation unit can be explicitly modelled with a possibility to use unit commitment within TIM.This model configuration includes the generation unit capacity, the fuel type of each generation unit, the start up costs (from cold, warm and hot), the efficiency curve of the generation unit to include startup and shutdown phases, startup and shutdown times (cold, warm and hot), the ramp rates, minimum load, efficiency at minimum load, minimum up-time and minimum down-time, the unit lifespan, the unit annual availability, and the start year of the unit (Geffert et al., 2018).The near-term generation unit pipeline is calibrated to EirGrid andSONI (2019, 2020) to account for early closures of units before their economic lifespan, which largely takes into account coal and peat based plants during the next decade.We have not forced on new capacity of future planned power plants from EirGrid and SONI (2020) in TIM, which includes recent battery storage installations.These recent and planned units can be forced on via a scenario file.
Existing 302 Onshore and 1 Offshore Wind farms within the Irish Wind Energy Association database currently operate as a single aggregated generation pool governed by historical hourly availability factors for wind generation from 2018.

Future Electricity system
New generation capacity investment costs, fixed and variable operation and maintenance costs and technical efficiencies are derived from Carlsson et al. (2014).The future unit commitment dispatch operational constraints and cycling costs are generalised for each technology type based on fuel and vintage (Kumar et al., 2012).
Renewable energy potentials are based on a number of sources.Total energy resource availability of solar and wind are from Future onshore wind energy potential has been assessed at high spatial resolution using GIS techniques.Wind farm expansion is constrained by both technology costs and a supply curve for grid connections based on a GIS analysis of existing houses, special areas of heritage and conservation, existing grid lines by voltage specification, wind energy potential, existing substation locations, and the costs per kilometre of standard ESB grid line specifications to connect wind resource potential to the existing grid.The wind energy potential is defined as a cost curve aggregated from 1938 suitable land parcels from a GIS analysis (Fig. 3).Due to the fact that increasing both technical details and temporal resolution causes exponential increase in model size, the spatial resolution is simplified and is represented in a 4 step cost curve (Table 3).This can be dissaggregated on a spatial resolution down to each individual parcel of land, or on a county by county basis.Future wind energy can generate at the same hourly capacity factor that occurred in 2018, however with capacity expansion.
Losses from electricity transmission and distribution are assumed to be 7%, with no grid expansion costs currently represented in the model.The maximum share of variable renewable energy including wind, solar and wave energy is constrained at 75% by 2030 and 100% is allowed by 2050.The model technology database includes short-term storage battery technologies, medium-term pumped hydro storage, and long-term hydrogen storage.All these storage technology options are typically deployed in hourly unit-commitment capacity-expansion scenarios and provide ancillary services and balancing in 100% VRES scenarios.EUR/tCO2 and an unlimited capacity.This caps the marginal abatement cost of the model at a price that does not exclude any of the plausible mitigation measures.

User constraints
User constraints applied to the Power sector in TIM largely pertain to the calibration of the near-term shutdown of coal and peat power plants before their economic life-time ends as a result of policy decisions.

Overview
The transport sector comprehensively describes the end-use transport technologies, and freight and mobility demands on a regional basis.This sector is divided into 26 counties across Ireland.To represent region-specific transport characteristics, some main parameters (vehicle fleet, transport infrastructure, fuel consumption, mileage, occupancy rate, load factor) are differentiated on a county level.Transport demand is split to three main categories: passenger, freight, and others.The passenger and freight demands are expressed as activity demands, and others are defined as a final energy demand (PJ).These final energy demands further split into aviation (international and domestic), navigation, fuel tourism and unspecified, aligned with the energy balance (SEAI, 2019)."Fuel tourism" refers to cross-border consumers and a portion of demand is used by unspecified modes.
The passenger transport demands are expressed in billion-passenger-kilometres (Bpkm).As shown in Table 4, the total passenger demand is divided into three classes of distance range including short-range (less than 5km), medium-range (5-30 km), and long-range (more than 30km) (NTA, 2018).Four transport modes satisfy travel demands: public services (bus, train, taxi), private cars, powered two-wheelers (PTW) and active modes (walk and bike).Non-motorised transport is only used for short-range trips, PTW are used for short-and medium-range travels, urban bus and school bus are used for shortand medium range travels, Intercity bus and heavy train are used for long-range trips, and light rail can only be used for the short-and medium-range trips in Dublin county.As shown in Fig. 4, demand for each mode can be met with a different set of technologies based on cost-optimisation and user constraints.The base year is calibrated based on the actual number of vehicles and the corresponding vehicle activities.Mode-specific relation between distance ranges is assumed constant throughout the modelling horizon (can be derived from Table 4).Total future passenger demand is obtained as a sum of mode-specific demand projections, described in subsections 3.3.3and 3.3.4.Table 6 shows vehicle characteristics on the national basis.
The inland freight demand is expressed in billion tonne-kilometres (Btkm).It comprises two main modes: goods trucks and train.The definition of light and heavy goods vehicles varies in different studies.In this model, they are disaggregated by three unladen weight bands: light-duty trucks (below 5 tonnes), medium-duty trucks (5-10 tonnes) and heavy-duty trucks (over 10 tonnes) (CSO, 2020g, h, i).Table 5 shows freight demand in the base year in million tonne-kilometres (Mtkm); the modal shares are assumed constant throughout the modelling horizon.slow growth during low income levels, rapid increase as income levels rise quickly and finally a saturation period.Gompertz 385 statistical model has been found to best fit the historical relationship between car ownership and income levels, although other functions have also been used in previous studies (Lian et al., 2018).The basic Gompertz function is shown in Equation 1.   with a CAGR of 0.7%.Passenger-kilometres of other modes (luas, train, small PSVs, and motorcycles) and active modes (walking and cycling) are projected using population as a driver.The passenger-kilometres are expected to increase by 60% with a CAGR of 1.5%.

International Aviation fuel demand
International aviation fuel demand is projected using number of passengers as a driver.The number of aviation passengers is projected using damped Holt Winters function based on historical time-series data obtained from CSO (Dantas et al., 2017;Grubb and Mason, 2001).The number of passengers in 2050 is expected to increase by 45.5% compared to 2018.The historical fuel demand for aviation and number of aviation passengers are then used as input for a linear regression model to project the future demand for aviation fuel.The fuel demand in 2050 increases by 37% relative to 2018 with a CAGR of 1%.

Other transport fuel demand
Demand for freight is projected using growth rates from AECOM (2019).The growth in tonne-kilometres of freight is expected to increase by 1.18 times in 2050 from 2018 level with a CAGR of 2.5%.Navigation fuel demand is projected using GDP as the explanatory variable.Fuel demand for navigation in 2050 is expected to increase 2.85 times compared to 2018 with a CAGR of 3.3%.Fuel tourism is assumed to remain constant at 11 PJ.

Future technology options
Common vehicle technologies and future options that are likely to become available for future investment shape the technology database for the transportation sector in TIM.They are categorised in five major groups (Aryanpur and Shafiei, 2015): 1. Internal Combustion Engines (ICEs), consists of spark ignition engines fuelled by gasoline, bioethanol, CNG, BioCNG, hydrogen and dual-fuel engines (running either on gasoline or CNG/BioCNG, each one taking 50% of the distance travelled), and compression ignition engines powered by diesel and biodiesel.
2. Hybrid Electric Vehicles (HEVs) are equipped with an ICE, which provides the main power, and a small electric motor to support the ICE and to recuperate the braking energy.
3. Plug-in Hybrid Electric Vehicles (PHEVs) have a similar powertrain to HEVs.Their batteries can be charged from the grid for driving tens of km solely on electrical power.We assume the maximum distance driven on electric mode is 50% in the base year and it can increase to 80% over time.
4. Battery Electric Vehicles (BEVs) solely rely on batteries, which provide the total motive power of the vehicle.The batteries charge from the grid.
5. Fuel Cell Vehicles (FCVs) are electrochemical devices that produce electricity through a reaction between hydrogen and oxygen.The electricity drives a vehicle's electric motor.Conventional fuel tanks are replaced with pressurised hydrogen storage tanks in FCVs.
Table A7 shows the techno-economic characteristics of future passenger transport vehicles (Mulholland et al., 2017;Helgeson and Peter, 2020).Maintenance costs are assumed to remain constant.Vehicle purchase price parity between BEVs and ICEs is expected in the period 2025-2030.
All these technologies compete to meet the mobility demand over the planning horizon.The model structure allows competition among stock replacement and fuel substitution within a mode.Modal shift may be simulated within each travel distance band by defining scenario-specific shares for the different modes (e.g.see Table 4).
Different fuels are supplied to the transport sector via four separated modules including supply, power, bio-refineries, and hydrogen modules.In other words, these connections integrate the transport sector with the entire energy system.
TIMES models usually use a simplified constant lifetime for different vehicles and thus, the vehicles are retired at the end of that lifetime.However, a detailed analysis of technology retirement profiles in Ireland shows that this simplified representation is far from reality (Mulholland et al., 2018).An actual profile shows a low decay in the beginning years and a long tail in the distribution over the longer term.To improve the retirement profile both for existing and new vehicles, TIM is equipped with realistic representation of the survival profile of car technologies.The survival rates are from the Irish CarSTOCK model (Daly and Ó Gallachóir, 2011;Mulholland et al., 2018).

User constraints
A set of constraints limits fuel and modal shares in transport sector as follows: -Maximum biodiesel share in passenger and freight transport demand is 4.1% in the base year.
-Maximum bioethanol share in passenger and freight transport demand is 3.2% in the base year.
-Maximum growth rate in new vehicle sales for advanced powertrains (HEVs, PHEVs, BEVs and FCVs) is 16% per year, which is enforced once the sales of a vehicle type reaches 15% of the total vehicle sales in the base year.
-The blend limit for biofuels is 10-12% for the regular ICEs without any modifications (i.e.10% bioethanol and 12% biodiesel with 90% and 88% gasoline and diesel, respectively).
-Total biofuel supply for the transport sector is allowed to double each decade, reflecting the rate of growth between 2010 and 2020 (NORA, 2021).

Energy service demands and projections
The residential stock projections up to 2040 are taken from Bergin and García-Rodríguez (2020) housing demand estimates, which utilise economic growth projections from Bergin et al. (2017).The stock is expected to increase by 40% from 2018 level with a CAGR of 2%.This results in an average of 27,600 new houses per annum between between 2021-2040.Beyond 2040, population is used as a driver to project housing stock.The total housing stock obtained in 2050 is 2.57 million dwellings which implies 8% increase from 2040.The filtered BER database is projected onto the total dwelling stock, using data from CSO (2020a, b) to calculate the total number BER ratings per archetype in the dwelling stock, as shown in Table 8.
BER assumes all buildings are heated to between 18°C for non-living areas (e.g.bedroom, bathrooms) to 21°C for living areas (e.g.sitting room, kitchen).This assumption is based on ISO 13790 calculations.To reflect actual energy use based upon internal temperatures in the Irish residential sector, the Archetype Dwelling Energy Model (ArDEM) (Dineen et al., 2015) was used to provide simulated annual energy consumption.ArDEM modifies the expected space heating energy consumption of each archetype and BER rating to the actual space heating energy consumption by adjusting internal temperatures in the building stock, a similar approach was used by Uidhir et al. (2020b).
The alternative internal temperatures assumptions in TIM are outlined in Table 9.After the base year, the change in the number of new dwellings per archetype drives demand, as previously outlined in section 2.5.All new dwellings mimic the energy intensity of the average base year A-rated dwelling for that archetype.This is 485 due to directive (EU) 2018/844, which requires all new residential dwellings to equate to at least a A2 BER by 2020.
The non-archetype energy service demands are: cooking, refrigeration, cloth washing, cloth drying, dish washing and electrical appliances.The non-archetype energy service demands are not dependent on the age, type or BER rating of the future housing type.These demands are projected to grow at the same rate as the growth in the total housing stock.
Non-Archetype energy demand data was obtained from SEAI (2018).This data was cross-checked with the JRC-IDEES (Integrated Database of the European Energy System ) database.The process involved cross-checking the share of residential energy service demands in TIM, against a database which had no input in to the model.

Future technology options
The TIM residential sector has two main mitigation options -switching heat technologies and fuels, and retrofitting existing dwellings to improve the thermal efficiency and uplift BER ratings.
While Ireland has ambitious retrofitting targets for 2030, data on the cost and energy savings for retrofitting is limited, particularly for deep retrofits.The expected costs and heat energy saving of retrofitting from one BER rating to another requires further investigation.The retrofit cost data is based upon AECOM (2013) and Ali et al. (2020) and the expected energy saving is based upon Collins and Curtis (2017).For this reason, the model uses a simplistic retrofitting options.There are two options for each archetype: shallow retrofit, which reduces dwelling space heating energy demand by 10-34%, and deep retrofit, which improves space heating energy efficiency by at least 35%.
The cost and expected heat energy savings per archetype by BER improvement is outlined in Table 10.The model currently implements a weighted average value, but a fully disaggregated retrofit could be implemented in a future model version.
The cost and efficiencies of new space heating and water heating technologies is sourced from Danish Energy Agency and Energinet (2020).The cost and efficiency of lighting, pumps & fans and non-archetype demands came from a range of sources including SEAI (2018) and Topten International Group (2021).

User constraints
-One space and water heating system per household (stoves excluded).This user constraint is applied to represent individual heating systems and calibrate results.
-Maximum 60% of existing buildings to be retrofitted by 2030, and 95% by 2070, the maximum value is interpolated 510 between 2030 and 2050.This user constraint is applied to align with maximum available labour.
-A prerequisite retrofitting requirement for electrical heat pump installation is applied to each archetype.Table 11 shows the percentage of dwellings which require a shallow or deep retrofit before an electrical heat pump can be installed (A "heat pump ready" dwelling must be at least a B2 energy rating).This user constraint is applied because of the lack of data regarding operation of electrical heat pumps in less than B2 rated dwellings, without this constraint heat pumps 515 would be installed in poorly rated dwellings and run at high performance.The constraint aligns with known heat pump performance data.-Maximum fuel share in cooking for natural gas and LPG is 40% and 10% respectively.Base year fuel share for natural gas is 32% and LPG is 1.5%.This user constraint is applied to align with the increasing market share of electrical induction hobs.

Industry
The industrial sector is modelled using a "top-down" methodology where energy demand is projected based on an assumed Energy demand for the industrial sector is projected using GVA per capita for each NACE category and population (Yakut and de Bruin, 2020).Historical energy consumption is obtained from SEAI (2019) energy balance.The total energy demand from industry in 2050 is projected to increase by 47% from the 2018 level.Cement demand up to 2025 is projected using Department of Finance ( 2020) stability program update which provides forecasts for the growth in modified investment.In 2019, 65% of the modified investment was in building and construction.Calculating a linear regression of the log of the index for output of the cement sector on the log of investment in building and construction at constant price results in an 18.6% increase in cement demand in 2025 from 2018 level.Beyond 2025, growth in cement demand is assumed to be the same as the growth in GNI*.This leads to a further increase by 17.8% between 2025 and 2050 at a CAGR of 0.7%.The energy intensity of the industry sector is expected to decline by 46.5% between 2018 and 2050 with a CAGR of -2%, reflecting historical trends.
Fuel switching is the only mitigation option available for combustion emissions from industry in the model.It is controlled through maximum predefined shares which are defined per year and subsector.Table 12 illustrates assumed maximum fuel switching shares which are possible across all the subsectors.Coal and coke to hydrogen 2% 6% 20%

Process emissions
Among the industrial processes, the cement industry has the largest share of process emissions, which is accounted for in the model.Historical data of cement production and emissions is obtained from the National Inventory Submissions (Duffy et al., 2020).Process emissions increased by 87% between 1990 and 2018 with a CAGR of 2.3%.The cement demand is projected using the number of new dwellings and the energy consumption of the corresponding industrial sector in a linear model.The amount of cement needed in 2050 is projected to double relative to 2018 with a CAGR of 2.2%.The demand for cement is then used to project process emission, which is expected to increase by 96% between 2018 and 2050 with a CAGR of 2.1%.

Services
The service sector in TIM comprises public and private services.It includes a representation of the following energy services: space heating, space cooling, water heating, cooking, refrigeration, building lighting and other appliances.Data centres electricity demand and public lighting are also represented.
Future fuel switching and technology-based efficiency options in the services sector are represented explicitly.However given a lack of sufficient building-level data to enable a detailed analysis, public and private services are modelled in an aggregated fashion (i.e. the building stock is not divided in categories).This is an area identified as a priority area for future model development.
The end-use demand services in the service sector include million meter square of area for public and private sector buildings, data centre demand and public lighting units.The area is projected assuming the same growth as in GNI*.The total area is projected to increase by 61.8% by 2050 from 2018 level with a CAGR of 1.5%.
Electricity demand for data centres is obtained from EirGrid's "steady evolution" scenario (EirGrid, 2017).The demand is expected to increase by 6 times in 2030 from 2018 level.We assume no growth in data centre demand after 2030 since permission requests for new/expanding existing capacities are not available yet.Further, post-2030 demand is even more uncertain given (i) the potential for exponential data usage growth and further data centres applications vs (ii) technology improvements and whether data centres fully utilise their contracted import capacity.Electricity consumption of data centres for cooling, and their potential to supply district heat from the excess heat they generate, are represented using the methodology described in Petrović et al. (2020).
Public lighting units are projected based on the Project Ireland 2040, whereby five major cities of Ireland, Dublin, Cork, Limerick, Galway and Waterford are expected to grow by 50% in 2040 (Government of Ireland, 2018).This results in a 12.5% increase in public lighting units in Ireland by 2040 from 2018 level, with a CAGR of 1%.Beyond 2040, the units are projected to increase by 1% per annum until 2050.

Agriculture
The current version of the agriculture sector in TIM comes from the Irish TIMES model and is documented in Chiodi et al. (2016).It includes representation of 12 energy service demands; half of them (dairy cattle, non-dairy cattle, sheep, pig, poultry and other animal rearing) belong to the livestock and half (production of pulses, potatoes, sugarbeet, barley, oats and wheat) to the tillage sector.Land availability and water consumption are explicitly represented and accounted for in the sector, however no specific constraints are set.Future energy service demands in the agriculture sector are left unchanged from the 2018 level.
Since the sector is a large and export-led part of the Irish economy, these should be adjusted based on a specific scenario narrative.
Crop-based bioenergy feedstocks are modelled within the agriculture sector.This includes growing of wheat, grass, rapeseed for biofuel production in the supply sector, as well growing of miscanthus, willow, and availability of forestry residues for biomass.Production of crop-based bioenergy feedstocks is constrained by potentials from SEAI (2015).
A representation of CH 4 and N 2 O emissions related to livestock and tillage is included in the agriculture sector.At the same time, as emphasised by Chiodi et al. (2016), due to the complexity of the sector, with a tendency for both technologies and productions levels to vary considerably among farms, challenges exist in identifying applicable mitigation options and their potentials.

Discussion & conclusion
This section discusses the strengths, limitations and development priorities for the TIMES-Ireland model, using the framework proposed by Pye et al. (2021) and DeCarolis et al. ( 2017), which outline best practice for ESOM development and priorities for new developments and applications of ESOMs for deep decarbonisation challenges.

Analytical advancements
The following is a summary of the main analytical advancements of TIM, explained in more detail throughout the text.
-The model has a flexible time-slice configuration with resolution possible up to the hourly level, which is necessary to e.g.model the power system under very high shares of variable renewables in deep decarbonisation scenarios, including power storage.
-The model also makes use of new TIMES features in the power sector, modelling dispatch, unit commitment and capacity expansion.
-The model has been developed with flexible regional definitions, with transport and power sectors detailed at the county level.
-Energy service demands are driven by a consistent set of population and macro-economic indicators, consistent with a national CGE model, and also disaggregated in sufficient detail to allow alternative demand scenarios, such as transport mode shifting, lower housing demand requirements etc.
-The development process strives to achieve best practice in software development, including version control, transparency (with open source and open data), quality assurance, documentation and wide stakeholder consultation.

Modelling for policy insight
The main purpose of the model is to meet the policy need to inform detailed sectoral pathways which can meet very ambitious decarbonisation targets.This subsection discusses TIM's strengths and weaknesses in this regard.
Firstly, the integrated whole system approach offer "macro systems" perspective.The temporal and spatial flexibility resolves the tension between the requirements for granularity for certain model functionalities, and the requirement for data and model tractability, and fast solving speed.A key strength of TIM is that it draws from openly-available and regularly-updated national datasets, which allows for future updates, and alignment with other models drawing from the same sources.
Secondly, an appropriately wide mitigation option space is required to meet deep decarbonisation challenges.Strengths of TIM include the rich depiction of important demand sectors, namely transport and residential, which are characterised to enable scenario variants of alternative demands.For example, splitting passenger transport demand into long-, medium-and shortdistance demand allows for switching to active and public transport demands .Similarly, the archetypal model underpinning the residential sector allows for lowering building floor area demand and internal temperatures, which can be key scenario variants.These scenario variants are exogenous, though there is increasing literature on endogenising behaviour in TIMES models.Several options for shifting to more efficient uses of energy, for example, are endogenous.However, the aggregate nature of the industry sector limits the potential analysis of demand reduction in this sector and is an area of future model development priority.
Another area for future model improvements is to explore carbon dioxide removal (CDR) options in more detail.At the global level, modelled pathways meeting the Paris Agreement rely on large-scale deployment of CDR options including biomass with carbon capture and storage (BECCS) (IPCC, 2018).BECCS is currently modelled within TIM, but options like Direct Air Capture (DAC) (Realmonte et al., 2019) are currently represented in an oversimplified manner; as real-world trials provide operational performance data, we will be able to improve the representation of DAC in TIM.
Another limitation of TIM is the sole focus on energy and process emissions.Agriculture and land-use are very significant emitters in Ireland, and future energy system decarbonisation trajectories will require a focus on the overlaps with other emitters.This is required firstly to understand bioenergy and waste potentials, and to take into account future gains from the circular economy, and also competition between energy crops, agriculture and reforestation.
A focus on the practice of model development and application is key.Robust quantitative analysis and information is an important ingredient in the energy and climate policy making process.But energy modelling should support the policy making process rather than determine its contents.This means the interface between energy modelling and the policy making process should be an iterative one that incorporates regular review and feedback as new issues and questions emerge (Strachan et al., 2016).TIM has been developed with the aims of openness and transparency, quality assurance and best practice in software development, and iterative engagement with key stakeholders to feed in sectoral and industry expertise, and to be able reflect policy and societal factors.Any model built to inform policy should be open to deep scrutiny early and throughout the process.It also requires capacity-development among the "consumers" of model outputs to ensure that results are interpreted, communicated and applied appropriately.
It is also important to communicate what the model can and can't do.For example, while TIM is a "cost-optimal" model, it does not capture many of the societal costs or benefits that the energy system imposes.For example, higher renewable energy shares may represent a higher overall investment and operation cost over the project lifetime, these may offer significant societal benefits in terms of lower air pollution, reducing the fossil fuel import bill and providing high-quality employment.Similarly, the distributional effects of the costs are not captured in the model, therefore considerations of a "just transition" are a challenge within TIM currently.

Future improvements
TIM will undergo continuous and iterative improvements and developments.In the near-term, the following is the priority list for future development.
-Modelling gas supply at hourly resolution -currently modelled at a seasonal level, which may unintentionally allow for energy storage capacity.
-Similarly, electricity interconnectors with the UK and France can be modelled at the same time resolution as the power sector, taking into account capacity constraints on top of annual trade constraints.
-Modelling hydrogen production routes other than from renewables (green hydrogen), such as blue, brown and grey hydrogen production options.
-A review and update of bioenergy conversion options, including an update of domestic bioenergy potentials.
-A review and update of future low-carbon technology costs.
-Developing further the Agri-TIMES module, first developed by Chiodi et al. (2016), in order to better represent agriculture, livestock and land-use emissions and mitigation options, including competition for land-use.
-A more detailed bottom-up focus on the industry sector, which is currently modelled in an aggregated top-down fashion, and disaggregation of the services sector.
-Developing a "Business as usual"/"With existing policy measures" case, which includes current policies and measures and hurdle rates to represent end-use technology uptake.

Conclusions
This paper presents the TIMES-Ireland Model, an energy systems optimisation model which is a significant step forward both in terms of analytical heft, open and best-practice development approach and in its contribution to national evidence-based policy development.TIM has been developed to help inform very ambitious decarbonisation objectives and to inform future carbon budgets, on the path to net-zero GHG emissions by 2050.
The model was developed from the legacy model, Irish TIMES, which has had a long history of contributing the the evidence base for Irish energy policy and for pushing the state-of-the-art in ESOM development.TIM also benefits from ongoing  Import of Ethanol 1st generation -Step 2 0.0 1.9 1.9 1.9 1.9 1.9 1.9 1.9

Figure 1 .
Figure 1.Simplified representation of reference energy system in TIM Pfenninger and Staffell (2016) andStaffell and Pfenninger (2016) respectively with availability factors fromRuiz et al. (2019).Ocean and tidal energy potentials are from O'Rourke et al. (2010), and wave energy power matrix 4 is derived from Nambiar et al. (2016).Hourly availability factors for ocean energy technologies are derived from the marine institute and OPW's data buoy services.

Figure 2 .
Figure 2. TIM hourly unit commitment modelling capabilities example for 2018

Figure 4 .
Figure 4. Transport structure in TIM

3. 3
.4 Passenger-kilometres: Other modes of transport Other modes of transport represent a much smaller share of mobility demand compared to private cars.Passenger-kilometres of large public service vehicles (PSVs) are projected using population as a driver in a log function and assuming average occupancy of 27.5.Large PSV passenger-kilometres are expected to increase by 24.2% in 2050 as compared to that in 2018

Fig. 5
Fig. 5 shows the RES diagram for the residential sector.Energy service demands are disaggregated between archetype and non-archetype demands."Archetype demands" are energy service demands which depend on the house type, namely the dwelling type and Building Energy Rating (BER) rating.The base-year residential energy demand by fuel is calibrated to the SEAI (2019) energy balance.Archetype energy service demands are particularly dependent on the type of building.The four energy service demands which are modelled based on archetype are: space heating, water heating, pump & fans and lighting.The residential building stock by type is explicitly modelled in TIM in three archetypes: detached, attached and apartments.The Archetype energy service demand data are sourced from SEAI (2020) BER database, which contains the raw data of 906,048 BER surveys.The BER database was filtered before use to remove outliers and any nonsensical values.The filters were based upon Dineen et al. (2015), Uidhir et al. (2020a) and group discussions within the National Modelling Network Ireland reducing the total records to 815,246.

Figure 5 .
Figure 5. Residential Reference Energy System future economic growth.Fourteen subsectors are represented and are based on SEAI (2019) energy balance.Baseline shares of energy carriers in the final energy consumption by subsector are assumed constant into the future and are based on the 2018 values (SEAI, 2019).
collaborations and interactions with other national models and data sources, including LEAP-Ireland, the ESRI's I3E model and SEAI databases.The model also benefits from the continuous development of the TIMES model generator within the 665 ETSAP community.Continuous, iterative and open model development is essential to ensure that the model remains fit-for-purpose and state-ofthe-art.The model has been built with these future calibrations and improvements in mind, with clear open documentation and development protocol to allow for ongoing improvements and updates, to better enable the Irish energy system to achieve the goal of carbon neutrality.

Table 2 .
Table2lists the energy service demands in TIM along with their corresponding drivers and values for 2018 and 2050.Specifics of the methodologies for projecting each energy service demand are detailed in later sections, for transport (3.3), residential (3.4), industry (3.5) and services (3.6).Energy service demands in TIM

Table 3 .
Onshore wind connection cost With this in mind, TIM has carbon capture and storage (CCS) technology options available from 2030, including retrofit options on existing coal, peat and gas power plants.Bioenergy carbon capture and storage (BECCS) is available in TIM from 2030 and provides net negative CO2 capture and allows negative emissions electricity generation.Direct air carbon capture and storage (DACCS) is also defined as a backstop technology i.e., it has a static unit cost of 2000 (Rogelj et al., 2018) potential locations 3.2.4CarbonCaptureandStorage, Carbon Dioxide Removal, and Negative Emissions technologiesIt is well documented in the literature that residual emissions are likely to remain in the future (net) zero carbon energy system from "hard-to-decarbonise" sectors(Rogelj et al., 2018).Carbon Capture and Storage (CCS), Carbon Dioxide Removal technologies (CDR), and Negative Emission Technologies (NETS) are currently seen within the literature as critical future technologies to capture the last marginal residual emissions to bring the energy system to net-zero CO2 or even net-negative 345 CO2 by mid-century.

Table 4 .
Total passenger demand and share of transport modes for each class of distance range(CSO, 2017(CSO,  , 2020c, e, f, g, h)   , e, f, g, h)

Table 5 .
Freight demand in 2018(CSO, 2020g, h, i)Future passenger car transport demand is projected based on future population growth and a growing rate of car ownership, which is in turn determined by income growth.Car ownership usually follows an S-shaped function which has three periods:

Table 8 .
Residential Dwelling Stock in 2018

Table 9 .
Internal Temperature Assumptions BER Rating Living Area Temperature Non-living Area Temperature A

Table 11 .
Prerequisite Heat Pump Requirement

Table 12 .
Maximum fuel switching shares

Table 13 .
Process emissions

Table A1 .
Import Fossil fuel commodity prices

Table A2 .
Import bioenergy commodity prices

Table A3 .
Import bioenergy delivery costs

Table A5 :
Base year generation units

Table A6 .
Residential Heating Cost & Efficiency.Note: it is assumed a water heaters are 30% less efficient than space heaters.The costs shown are for space heaters only.

Table A7 .
Techno-economic characteristics of passenger transport vehicles