A multi-layer energy modelling methodology to

27 downloads 0 Views 886KB Size Report
Jan 1, 2019 - electricity integration strategies: the case of the residential cooking sector in Italy .... Models for the generation of load demand profiles represent a critical ... commercial and light industrial user types, with an half-hourly resolution ..... least in terms of average total energy produced per family on a daily basis.
Accepted Manuscript A multi-layer energy modelling methodology to assess the impact of heatelectricity integration strategies: the case of the residential cooking sector in Italy

Francesco Lombardi, Matteo Vincenzo Rocco, Emanuela Colombo PII:

S0360-5442(19)30004-0

DOI:

10.1016/j.energy.2019.01.004

Reference:

EGY 14477

To appear in:

Energy

Received Date:

29 October 2018

Accepted Date:

01 January 2019

Please cite this article as: Francesco Lombardi, Matteo Vincenzo Rocco, Emanuela Colombo, A multi-layer energy modelling methodology to assess the impact of heat-electricity integration strategies: the case of the residential cooking sector in Italy, Energy (2019), doi: 10.1016/j.energy. 2019.01.004

This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

ACCEPTED MANUSCRIPT

A multi-layer energy modelling methodology to assess the impact of heat-electricity integration strategies: the case of the residential cooking sector in Italy Francesco Lombardia, Matteo Vincenzo Roccoa, Emanuela Colomboa aPolitecnico

di Milano, Department of Energy, via Lambruschini 4, Milan, Italy

Corresponding author: Lombardi F, Tel.: +39-02-2399-3866; address: Via Lambruschini 4, 21056 Milan, Italy. E-mail: [email protected]

Abstract To support the ongoing transition towards smart and decarbonised energy systems, energy models need to expand their scope and predictive capabilities. To this end, this study proposes a multi-layer modelling methodology that soft-links (i) a stochastic bottom-up load curves estimation model, (ii) a technology-rich energy system optimisation model (Calliope) and (iii) a Multi-Regional Input-Output model (Exiobase v.3), and applies it to investigate the economic and environmental consequences entailed by a massive replacement of traditional gas-fired kitchens with induction kitchens within the Italian residential sector. Two scenarios are considered for the analysis: (i) business as usual (BAU, 2015 energy system configuration), and (ii) national energy strategy (SEN, configuration prospected in 2030). The results show how the intervention produces positive net effects on the primary energy balance of the energy sector only when sustained by adequate shares of renewables, as in the SEN (-1.5 TWh∙y-1); otherwise, increased operation of fossil-fuel plants offsets gas savings (BAU, +2 TWh∙y-1). Nonetheless, feedbacks on other productive sectors entail additional energy consumption and emissions, thus counterpoising positive effects obtained within the energy sector even in the SEN scenario. Still, higher renewables penetration reduces overall additional emissions from 2.07 Mton∙y-1 for BAU to 0.88 Mton∙y-1 for the SEN.

Keywords: Energy modelling; Input-Output analysis; Electrification pathways; Cooking devices; Integrated Assessment Models; Heat-electricity integration

ACCEPTED MANUSCRIPT 1. Introduction Policies for the decarbonisation of energy systems are registering an increasing attention and rate of implementation worldwide, and especially at the European level, as a result of the regulatory framework that each country set out to comply with the COP21 commitments [1]. While efforts have been primarily focusing so far on the deployment of renewable power capacity within the electricity sector [2], most recent researches are showing how coupling multiple energy sectors (e.g. electricity, heat, transport) into a smart energy system concept is an essential prerequisite to identify cost-optimal energy system configurations that would not be achievable within a traditional single-sector perspective [3]. To this regard, Bloess et al. [4] report that, among studies comparing alternative sector-coupling options, there is a consistent agreement that heat-electricity integration is the most cost-effective strategy and the one that should be prioritised. Indeed, heat – here considered as the sum of space heating, water heating and cooking energy uses – accounts for about 50% of total final energy consumption within the EU28 region [5] and for up to 85% of residential final energy consumption [6], while its supply is largely dominated by non-renewable sources. On the one hand, the simple electrification of part of the heat supply allows for the deployment of technologies, such as induction kitchens and heat pumps, which ensure much higher energy conversion efficiencies than their combustion-based counterparts [7]. On the other hand, the integration of heat and electricity allows for the use of smart thermal energy storage, which represents a low-cost option to store excess renewable electricity [8] and to increase the overall system flexibility and the renewables penetration [4]. This study focuses on the case of the Italian cooking sector, which is currently dominated by natural gas and LPG, significantly lagging behind the EU average in terms of penetration of efficient electric appliances: as shown in Figure 1, up to 80% of total energy consumption for residential cooking in Italy is supplied by fossil fuels (70% natural gas and 10% LPG), compared to an EU28 average of about 46%. The latest Italian National Energy Strategy (SEN 2017) [9] aims at increasing the penetration of efficient induction kitchens in the residential cooking sector in the short term, in compliance with a broader strategy of progressive electrification of the heat sector, traditionally grounded on natural gas. However, the impact that a massive electrification of households’ kitchens may have on the power sector (e.g. change in power plants operation due to a change in electricity demand, as recently shown by the IEA for the case of the electrification of the transport sector [10]), on the broader economy and on the environment has not been yet quantified. In general, identifying optimal power system configurations and decarbonisation pathways within an integrated heat-electricity system poses some challenges. In fact, despite the recent developments in the field of opensource energy modelling frameworks (such as: Calliope [11], oemof [12], Dispa-SET [13] and PyPSA [14]) added to the literature with the required modelling features in terms of spatial and temporal resolution and capacity of handling multiple energy carriers [15], there is still a significant lack of complementary models for the generation of demand profiles, particularly heat demand profiles [16], which are an essential prerequisite to perform integrated analyses. Secondly, energy modelling frameworks commonly consist of technology-rich bottom-up representations of the energy sector alone, whilst policy interventions on the latter entail economic and environmental consequences on the whole set of productive sectors within a national economy (e.g. modifications in the energy-sector supply chains, need for the manufacturing sector to produce new technologies, indirect GHG and pollutants emissions, etc.). Such consequences need to be taken into

ACCEPTED MANUSCRIPT account as well in order to avoid the provision of incomplete and misleading information to policy makers [17]. This is particularly relevant when trying to assess heat-electricity integration strategies, as these often involve changes in the existing supply-chain (e.g. a shift from gas to electricity) and the introduction of new technologies (e.g. heat pumps, electric boilers, induction kitchens) that require additional resources and emissions to be manufactured. Though Integrated Assessment Models (IAMs) are commonly employed as a means to enlarge the domain of energy policy analyses in this direction, they allow doing so only at the expense of a simplification of the energy system characterisation, which cannot ensure anymore the level of detail that is required to reproduce the operation of modern energy systems [15]. Recent studies propose alternative methodologies to achieve similar results without sacrificing the detail of the energy system representation, such as soft-linking bottom-up technology-rich energy models with top-down macroeconomic models [18].

Share of fuels in the cooking sector

100%

4.2%

4.6% 15.5%

80% 49.2%

9.8%

60% 40% 20%

12.8%

Renewables and Wastes Solid fuels Electricity Oil & petroleum products Natural Gas

70.1%

33.1%

0%

EU-28

Italy

Figure 1. Share of fuels in the cooking sector, comparison between EU28 average and Italy (Source: Eurostat [6]).

This paper aims at contributing to the overcoming of the abovementioned modelling challenges, by providing a general approach for soft-linking bottom-up energy models with top-down Input-Output models. More specifically, this study sets up a multi-layer energy modelling methodology that allows to evaluate alternative heat-electricity integration strategy, combining the following three steps: 

Quantification of the additional electric load required by a replacement of fossil-fuelled heat technologies with electrically-powered alternatives with a high temporal resolution.



Identification of the effect of such additional load on the cost-optimal operation of the energy system, i.e. on the cost-optimal mix of electricity production.

ACCEPTED MANUSCRIPT 

Assessment of the economy-wide economic and environmental effects associated with the changes introduced in the energy sector and with the ancillary activities of the background economy that are required to support the analysed transition.

The proposed approach is then applied to assess the economy-wide consequences entailed by a massive replacement of natural gas kitchens with efficient induction kitchens in Italy. The impact of this intervention is assessed by considering two different national power system capacity mixes: the current scenario assumes the installed capacities of power technologies of the year 2015; the renewables scenario assumes the installed capacities defined as per the 2030 goal by the Italian energy strategy. This comparison is useful to verify the synergistic potential of electrification technologies and the deployment of renewables. This study adds to and extends the current literature in several ways: first, it proposes a comprehensive softlinking approach, which enables to assess the overall impact of a technology transition without sacrificing the technical detail of the energy modelling layers. Secondly, it includes in the analysis an additional and critical layer related to the detailed assessment of changes in load demand profiles, for which an ad-hoc model is developed. Finally, it provides a general modelling methodology that may be applied consistently to further heat-electricity integration studies. The practical application completes the work by demonstrating the potentiality of the methodology for an unexplored case study, relying on open-source models for all the three layers. The article is organised as follows: a review of recent studies is provided in section 2; section 3 presents the proposed soft-linked approach, which is applied for the analysis of the Italian case study in section 4; results and discussion are collected in section 5. Concluding remarks are finally formulated in section 6.

2. Literature review This section reviews the recent studies related to (i) country-scale load profile assessment; (ii) energy system dispatch optimisation models; (iii) economy-wide impact assessment models.

2.1. Country-scale load profile assessment Models for the generation of load demand profiles represent a critical prerequisite to investigate energy scenarios that involve changes in energy end uses, such as the introduction of new residential electric appliances, the electrification of previously non-electric energy uses, or demand-side management strategies [19]. The issue is particularly relevant for the case of heat-electricity integration studies, as current heat demand profiles are rarely available in the form of historical time series due to the lack of load meters in most contexts, especially those where heat is supplied through country-scale gas networks and for multiple end uses (e.g. space heating, water heating, cooking) at once, as for the case of Italy [20]. A few models or approaches for the assessment of country-scale load profiles have been proposed in the literature, typically revolving around the bottom-up concept of obtaining heat demand profiles for individual building archetypes in different locations and then aggregating those based on considerations about

ACCEPTED MANUSCRIPT population and climate data within the identified locations. For example, Calise et al. [20] generate hourlyresolution residential space heating, water heating and cooling profiles for multiple Italian regions, by relying on dynamic building archetypes modelled through the commercial software TRNSYS, and subsequently aggregating them by taking into account population density and building type distribution across the national territory. Similarly, Clegg and Mancarella [16] derive space heating and water heating profiles for residential, commercial and light industrial user types, with an half-hourly resolution and for 404 areas across Great Britain, by relying on the software EnergyPlus for the generation of individual building-archetype profiles. A further example adopting a similar logic but relying on empirical data for individual load profiles, rather than on simulation software, is represented by the open-source model demandlib [21], developed by the Reiner Lemoine Institut and making use of a sigmoid function based on German building datasets to simulate the response of different building archetypes to outdoor temperature changes. Converely, the DESSTInEE model, proposed by Bobmann and Staffell [19], allows to generate future electricity demand scenarios by changing the relative diffusion and efficiency of several types of appliances and technologies within existing electricity load curves for most European countries, including some heat-generating appliances. Nevertheless, none of the reviewed approaches focuses on the assessment of cooking load curves, nor includes it within the domain of the heat demand analysis. Unlike other heating loads, residential cooking is not dependent on outdoor temperature, but individual cooking profiles are nonetheless highly variable in terms of power absorbed, timing and duration, depending on the users’ cooking habits (e.g. typical meals prepared, cooking style preferences) and context-specific cooking culture. Accordingly, it may be still possible to generate those by means of a bottom-up approach aggregating individual cooking profiles, in this case modelled for several user archetypes rather than buildings; however, the natural variability of cooking tasks should be properly considered to differentiate between each user’s behaviour.

2.2. Energy system dispatch optimisation The number of energy system models and modelling frameworks has seen a significant rise in recent years [22], as a consequence of the inadequacy of most long-time established alternatives to cope with the requirements of modern energy systems operation. Indeed, the increasingly high shares of variable renewable sources and of decentralised generation within modern energy systems led to the rise and to the current predominance of bottom-up energy models characterised by a high spatial and temporal resolution [15]. In particular, a significant contribution to this trend has been brought by models and frameworks arisen within the Open Energy Modelling Initiative (OpenMod) [23], which also contributed to stress the importance for openness and transparency of both models and datasets [24,25]. The various state-of-the-art models differentiate in terms of their specific focus and range of applicability, yet the most relevant and widespread are versatile enough to serve different purposes (such as, for instance, optimising both the investment planning and the operational strategy of an energy system). For instance, the Open Energy Modelling Framework (oemof) [12] developed by the Reiner Lemoine Institut can be used both as an energy planning and an energy dispatch optimisation model. The same is valid for PyPSA, developed by Brown et al. [14], though with a more specific focus on power system analysis. Another remarkable

ACCEPTED MANUSCRIPT example is represented by Dispa-SET, developed by Quoilin et al. [26] within the European Joint Research Center (JRC) to model and optimise the operation of national and cross-national energy systems, and already applied by the same authors to various case-studies [27,28]. A peculiar case is that of EPLANopt, developed by Prina et al. [29] by adding an open-source evolutionary algorithm for multi-objective optimisation to the existing and well-known simulation model EnergyPLAN, and thus making it suitable for hourly dispatch optimisation analyses. Finally, Pfenninger [30] developed the multi-scale open-source modelling framework Calliope, allowing to optimise both dispatch strategy and investment planning and capable of handling with an easy formalisation multiple locations or regions and multiple energy sectors, as demonstrated with a case-study for the UK [11]. For further examples, the reader is referenced to Ringkjøb et al. [22] and to the OpenMod database [31]. Each of these models has its own peculiarities and strengths, and the best choice essentially depends on the modellers needs; also, the open-source nature and the conception of most of these models as open frameworks leaves large possibilities for the modeller to add any additionally-required feature.

2.3. Economy-wide impact assessment While bottom-up approaches are usually based on analytical models of the analysed technologies, top-down models mostly rely on empirical datasets derived from Leontief’s Input Output tables [32]. For this reason, application of simple and linear top-down models is usually referred as Input-Output analysis (IOA). Even if the scope of top-down models is comprehensive, such models are characterised by a high aggregation level: indeed, energy technologies are usually lumped together in one average “energy sector”. For such reasons, the two approaches have complementary rather than opposite features, and this invites in developing methods for their joint use, usually called “links”, which are increasingly proposed in the recent literature [7]. In this perspective, optimisation models can be hard-linked or soft-linked. Hard-linked models consist in a mathematical merge of the two models into one single model: Jacobsen [33] adopts this approach to assess the effectiveness of financial and technical instruments to reduce GHG emission in Denmark in a general equilibrium framework. Likewise, Bauer et al. [34] propose the REMID-R model, including energy, economy and climate models to assess the effect on public welfare associated with the introduction of renewables in the energy mix. Based again on market equilibrium mechanisms and on a detailed technology characterisation, the PRIMS model has been used in several studies to address the transformation of the European energy system and the effectiveness of several environmental policies [35]. Gargiulo and Gallachóir [36] present a detailed descriptions of other hard-linked models, such as MERGE and POLES, specifying their capabilities and limitations. As an alternative to the hard-link approach, a soft-link usually consists in solving separately the top-down and bottom-up models, linking them through endogenous and exogenous variables in a closed loop that iteratively return a unique solution. In its most simplified form, the soft-link can be established in two ways: 

Linking Top-down to Bottom-up. In this case, the focus is still on the energy sector only, and the topdown model works like an LCA model, assessing the indirect environmental consequences caused

ACCEPTED MANUSCRIPT by the energy sector in future energy scenarios. Recent literature related to this approach is investigated in a number of studies by the Authors [37,38]. 

Linking Bottom-up to Top-down. In this case, results of the former are introduced as an input for the latter: hence, the focus is now on the whole economy, assessing the economy-wide consequences due to future energy scenarios. The present work proposes this second type of soft-link and it has been applied by the Authors in a recent study [39].

Messener et al. [40] proposed a soft-link between MESSAGE and MACRO models, with the aim to assess the impact of energy supply costs on the national energy production mix in a general equilibrium framework. Kober et al. [41] linked a nonlinear macroeconomic model to an energy system model by considering the decreases in consumers’ spending due to the introduction of carbon taxes. Recently, Heinrich et al. assessed the impacts on the German economy of a decommissioning of coal power plants through a softlinked model: they revealed that the proposed phasing out is not sufficient for Germany to reach its target level on GHG emissions, and they highlighted the relevant role of indirect GHG emissions caused by renewables and the related infrastructures [42].

3. Methods and models This study proposes to set up an innovative multi-layer modelling methodology, which encompasses all the relevant aspects of a given heat-electricity integration strategy (as they have been defined in sub-section 2.1) and allows to analyse each of them in detail as a unique modelling layer, as schematically presented in Figure 2.

Figure 2. Conceptual representation of the proposed multi-layer modelling methodology.

ACCEPTED MANUSCRIPT The first modelling layer (Layer 1) consists of the bottom-up synthetic generation of load demand profiles, which, in the framework of heat-electricity integration, may include a partial or complete substitution of energy carrier uses. With reference to the specific case-study of the Italian cooking sector, this first layer of modelling implies the bottom-up computation of the additional cooking load that would be charged onto the current electricity load curve if all residential gas kitchens were to be replaced by induction kitchens. Due to the absence of available datasets or models for the generation of cooking loads, a novel and case-specific model is built within this study, adopting a bottom-up stochastic approach that allows to consider both usertype differences and unpredictable random user preferences (see technical details in sub-section 4.1). The second layer (Layer 2) consists in the technology-rich and highly-resolved representation of the energy system, of which the previously modelled load demand constitutes an essential input element. At this level, for a given energy system configuration, the cost-optimised dispatch strategy required to supply the load is obtained. For this study, the capabilities of existing open-source modelling frameworks allow for the required technological characterisation: Calliope is selected due to its high degree of internal validation, reliability, completeness of documentation and transparency. Finally, to assess how the investigated change in the energy system operation and the introduction of new appliances do impact on the broader economic and environmental systems, the modelling Layer 2 is softlinked to a Multi-Regional Input-Output model (Layer 3): the Exiobase v.3 [43,44] is adopted for this study. This final modelling stage allows to account for the prospected economy-wide consequences of the analysed shocks, focusing on the economic impact (proxied by the sectoral value-added generation) and environmental impact (proxied by the sectoral CO2 emissions). The proposed approach ultimately allows to evaluate the overall effects on the national economy due to a change in the operative conditions of the power sector, taking advantage of a detailed and highly-resolved representation of the latter.

4. Case study: electrification of Italian cooking devices This section describes how the methodology outlined in section 3 has been applied to the selected case study, consisting in the total replacement of currently gas- or LPG-fired residential kitchens with electricallypowered induction kitchens. Two different scenarios have been defined: 

Business As Usual (BAU). In this scenario, the replacement of cooking devices takes place while the installed capacities of power technologies are those available in Italy in 2015.



National Energy Strategy (SEN). In this scenario, the intervention takes place within an energy system configuration characterised by the share of variable renewables forecasted by the latest Italian energy strategy for the year 2030. This choice is dictated by the fact that the capacity of VRES to contribute to the supply of the simulated additional electric cooking load can significantly affect the total primary energy supply (TPES) balance, and thus the effectiveness of the analysed electrification strategy.

For both the scenarios, results for a baseline case (without the additional cooking load) and a shocked case (with the introduction of electric cooking) are derived and comparatively assessed. Moreover, both scenarios

ACCEPTED MANUSCRIPT assume the electricity demand profiles of 2015 and a same initial penetration of induction kitchens in Italian households (assumed at 5%), meaning that up to 23.38 million of kitchens are interested by the intervention [45]. All the models adopted or specifically developed for this study are based on the Python language; newly developed models and all the input data used throughout the study are shared as open-source material at the GitHub repository “SESAM Polimi – Italian cooking sector paper” [46].

4.1. Load curves estimation A case-specific model is adopted for the generation of synthetic electric cooking load curves, which are subsequently added to the baseline Italian electricity demand curve – obtained from Terna [47] and ENTSOE [48], as elaborated by the Open Power System Data portal [49]. Figure 3 shows the logical flow defined for the bottom-up stochastic computation of country-scale cooking load profiles, which is repeated for every daily aggregate profile for a total of 365 runs. Firstly, two main user archetypes are defined, namely small and large families, comprising households of 1-3 members and 3-6+ members, respectively. Based on ISTAT data [45], this corresponds to a total of 18’264’150 small families and 5’117’028 large families. Each user archetype is then further differentiated in terms of meal-specific behaviour. In fact, reports from Federazione Italiana Pubblici Esercizi (FIPE) and Confcommercio [50] provide information about the average shares of households that consume a given meal – data are differentiated per breakfast, lunch or dinner – out of home (i.e. in restaurants, bars, etc.) with a given frequency. Six different frequency classes are identified for each meal type. With such information, it is possible to simulate each meal for each family within each user archetype by also checking the probability of the given meal not being cooked at home and thus not entailing the switch-on of the induction kitchen. Only when the meal is cooked at home, the model simulates an induction kitchen switch-on, giving rise to a cooking event: the event is simulated uniquely for each individual household considered, and each event entails a stochastic variation of the timing, duration, power level and cooking cycle1 shape within some pre-defined ranges. The two main meals (lunch and dinner) are modelled as the combination of two cooking events, with the idea to simulate the contemporary preparation of a main dish and a side dish, or the use of two different cook plates for different ingredients of a unique dish. Time-related parameters are defined with a 1-minute time resolution to keep a high adherence to real-life duration of cooking events, and only later averaged on an hourly basis to allow merging with the hourly-resolved historical electricity load. Further details about the data and the assumptions adopted are reported in Appendix 1. In particular, the exogenous parameters required by the model to perform such workflow are collected in Table 3: such inputs are elaborated based on the previously-mentioned references by FIPE, Confcommercio and ISTAT as well as on a specific report by CENSIS and Coldiretti [51], and in

1

“Cooking cycle” is here intended as the sequence of tasks required for the preparation of a given meal. For instance, the preparation of

a pasta requires a first high-power task to bring a pot of water to boil, and a second lower-power task in which the water is kept around boiling temperature and pasta in boiled for a pre-defined amount of time.

ACCEPTED MANUSCRIPT some cases (e.g. timings required to boil water at different power levels) with the support of the Cook-STePS software [52].

Figure 3. Schematic representation of the bottom-up stochastic model defined for the computation of aggregate cooking load profiles. Despite the lack of empirical datasets against which to compare the shape of the loads generated by the model, it is possible to combine data from the IEA [53] and from the Eurostat [6] to obtain an average figure of 2.228 kWh∙household-1∙day-1 of gas consumed for cooking at the residential level in Italy, or 19017.1 kWh∙year-1 at the aggregate level; taking into account also the different cooking efficiencies estimated by the US Department of Energy [54] for gas and induction kitchens, the value can be used to validate the model at least in terms of average total energy produced per family on a daily basis.

4.2. Energy system model application The Calliope energy system optimisation model has been set and calibrated based on a single-node representation of the Italian power system. As demonstrated by Prina et al. in two applications [55,56], results of this assumption are undistinguishable from a multi-node representation for a penetration of VRES compatible with that assumed for both scenarios. The main exogenous parameters provided to the model are: 

Electricity load demand profiles. The baseline electricity demand profile is obtained from freely available datasets for the year 2015 with hourly resolution. Since the change in electricity demand is given by the increase in consumption of induction kitchens, the modified profile is obtained by adding to the former the newly generated cooking load profiles, estimated as described in sub-section 4.1.



Yearly hourly VRES time series. These are obtained in the form of capacity factors from the Renewables ninja database [57,58], and consistently applied for the two scenarios.



Installed power capacities. The energy model provides the cost-optimal strategy for electricity production and dispatch with a rolling-horizon strategy: therefore, two fixed sets of installed capacities are defined (one for each scenario), and exogenously provided to the model (Table 1). In particular, installed capacities for the SEN scenario are modified so as to reflect the short-term

ACCEPTED MANUSCRIPT energy system evolution prospected by the SEN: CCGTs are the only operative fossil fuel technology, in compliance with the ongoing strategy to phase out all coal and oil plants, while VRES capacities correspond to those prospected by the SEN for 2030 [9]. 

Power technologies data. Technical parameters (e.g. efficiencies, ramp factors, etc.) and economic costs are gathered from RSE [59] and Terna [47] for the reference year 2015.

Some additional constraints are introduced within the model formulation so as to take into account that the actual dispatch of some of the installed power plants – such as those about to be dismissed (i.e. traditional coal, oil and other fossil fuels plants) or those dependent on natural parameters (i.e. hydro reservoirs and run-of-river plants) – is indeed limited by non-technical and non-reproducible factors [59]. Hence, the maximum dispatchable capacity per time-step of those plants is bounded so as to comply approximately with the total annual energy generation reported by Terna [47]. In addition, ramp constraints are introduced for CCGTs and other fossil-fuel power plants to better reflect technical limits in plants dispatch.

Table 1. Summary of installed generation capacities (MW) and storage capacities (GWh) for the BAU and SEN scenarios. BAU MW (GWh)

SEN MW (GWh)

47630.8

47630.8

Coal (traditional)

6855.2

0

Coal (USC)

1845.0

0

Oil & other fossil

6053.0

0

Waste-to-energy

8351.9

8351.9

Power technology capacity Combined Cycle Gas Turbine (CCGT)

Geothermal Wind PV (farm)

768.0

768.0

9137.0

9137.0

7000.0

21913.0

11892.1

37277.4

Hydro reservoir and basin

9425.0

24365.3

Hydro run-of-river

5203.3

5203.3

Biomass (solid)

575.7

575.7

Biomass (biofuel)

909.8

909.8

Biomass (biogas)

1233.2

1233.2

PV (rooftop)

Import capacity Pumped-hydro-storage

6740.0

6740.0

7591.8 (700)

7591.8 (700)

To check the model consistency, the annual mix of electricity production resulting from the energy system configuration for the BAU scenario (without the addition of cooking loads to the electricity demand) is compared to the electricity mix of 2015 provided by Terna for the Italian liberalised electricity market. For most sources, the quantitative difference between model results and Terna results at the end of the year is lower than 5%. Higher discrepancies are obtained, as expected, for photovoltaic (as photovoltaic power is, in reality, continuously installed throughout 2015, while we assume all capacity being available immediately in the model), bioenergy (subject to incentives for cogeneration which are not taken into account here) and

ACCEPTED MANUSCRIPT geothermal (for which we assume constant production, though actual plants have availability factors around 90%). Considering the limited contribution of the latter two, the overall production trend demonstrates a high legitimacy compared to the real behaviour, at the light also of the importance of avoiding overfitting with historical data [60].

4.3. Soft-linking energy and Input-Output models Given an economy composed of n sectors, each with s types of exogenous transactions (say, primary energy, GHG emissions, value added generation, etc.), l energy technologies, and considering a time frame of one year, the Input-Output model is grounded on the operators defined by equation (1).

AN

A=

 CD

CU  f N   ; f =   ; b = b N b E  A E  f E 

(1)

The technical coefficients matrix 𝐀(𝑛 × 𝑛) expresses the technology linkages between all the national sectors in the economy. The final demand vector 𝐟(𝑛 × 1) collects the final expenditures (mainly households, govern and investments). The exogenous transactions coefficients matrix 𝐛(𝑠 × 𝑛) represents the value-added generation, the direct resources consumptions or waste emissions of each sector per unit of its product. Notice that these operators can be referred to the energy-related sectors (subscript “E”) and to all the other national sectors (subscript “N”). M matrices 𝐂U(𝑛 ― 𝑙 × 𝑙) and 𝐂D(𝑙 × 𝑛 ― 𝑙) are respectively the Upstream and Downstream Cutoffs related to the energy sector: for each energy technology, 𝐂U collects inventories of national products yearly required to support its production, while 𝐂D represents the amount of electricity delivered to all the other national sectors for each unit of their production. All the data presented in equation (1) are derived from the Exiobase v.3 database (http://www.exiobase.eu/) [44], which collects the monetary multi-regional input output tables covering years 1995-2011 for 49 world economies (of which 5 Rest of the World regions), 163 industries per economy (of which 12 disaggregated energy technologies) and several environmental extensions. The proposed soft-link procedure is described by the following phases: 

Phase 1: electricity production mix update. Before applying the policy shock, it is necessary to update the electricity production mix of the Exiobase database from 2011 to the reference baselines (superscript “0”), that is, the energy mix of 2015 for the BAU scenario, and the prospected energy mix in 2030 for the SEN scenario. This is performed by updating the downstream cut-off matrix 𝐂𝐃→ 𝐂0𝐃 and the energy final demand vector 𝐟𝐄→𝐟0𝐄 according to equation (2), where 𝐢(1 × 𝑙) is the summation vector and 𝐸0𝑖 is the production of energy by the i-th technology. Notice that only the shares of energy production technology are modified, while the overall energy production yield remains constant.

 E0







C0D =  i 0   i  CD   Ei i

;

 E0







f E0 =  i 0   i  f E   Ei i

(2)

ACCEPTED MANUSCRIPT 

Phase 2: application of the prospected policy shock. Once the Input-Output model in the baseline year (0) has been fully characterised, the prospected policy intervention is numerically applied to the model, and the overall impact ∆𝐁(𝑠 × 𝑛) is accounted through equation (3), where 𝐈(𝑛 × 𝑛) is the identity matrix. The policy shock is implemented in the model according to the ceteris paribus principle, that is, by assuming that the only changes introduced in the baseline economy (0) consist in the following three overlapping effects: 1. Change in households’ final demand of natural gas, represented by the reduction in economic consumptions ∆𝐟𝐍𝐆 and the related avoided CO2 emissions ∆𝐁𝐍𝐆. These data are provided by the load curves estimation process (sub-section 4.1): the expected avoided natural gas consumption is about 18738 GWh, with an average CO2 emissions factor of 55 ton∙TJ-1. 2. Change in households’ demand for electricity ∆𝐟𝐄𝐄: these data are provided by the energy optimization model (sub-section 4.2). Since the increase in electricity production is all required for households’ cooking needs, it is assumed that the background electricity mix remains unchanged and equal to the baseline (0). 3. Increase in economic production in the manufacturing sector ∆𝐟𝐈𝐂 that is required to produce induction cookers. The number of cookers to be substituted is about 23.4 million, and an average unitary cost of 450 € each is assumed, with an average lifetime of 10 years. The investment cost is allocated 50% in the Manufacture of ceramic goods sector and 50% in the Manufacture of machinery and equipment n.e.c. sector. For the sake of simplicity, it is assumed that the whole manufacturing of the induction systems is performed in Italy: this simplification is useful to understand the overall impact of the cookers manufacturing that would be much difficult to quantify if split among the real producers.

 B =  b0N 

b

0 E

  A 0N    I   0   CD 

 C0U      diag  f NG  f EE  f IC    B NG  A E0    1

(3)



The application of equation (3) returns the impact of the analysed intervention in terms of value-added generation (i.e. the Gross Domestic Product, calculated from the revenues side) and CO2 emissions, each distinguished by country and by economic sector of origin. In general, an increase in CO2 emissions can be immediately interpreted as a negative effect of the designed policy. An increased value-added generation can be interpreted as a negative effect as well, since it represents an additional economic effort required to obtain the same useful effect (i.e. the amount of meals cooked does not change before and after the implementation of the induction kitchens). Finally, it is worth noting that the applied Input-Output model does not include any market or social dynamic mechanisms such as price equilibrium or rebound effects, which may represent an interesting future development of the current work.

ACCEPTED MANUSCRIPT 5. Results and discussion The load curve estimation process introduced in sub-section 4.1 returns stochastically-generated daily cooking load profiles with a 1-min resolution, representing all the households interested by the intervention. Figure 4 (left side) shows 30 stochastic profiles and 1 average profile, where the absolute peak load is represented by dinner cooking and is comprised in a range between 8 and 10 GW, depending on the outcome of the stochastic algorithm for a given day. Figure 4 (right side) shows the overall load curve for a sample week, where peak periods of the additional load associated with lunch and dinner cooking approximately overlap with the baseline load peaks, thus giving rise to a significant increment of the overall absolute peak load.

Figure 4. Left side: example of stochastically-generated aggregate induction cooking load profiles for 30 days (thick line represents the average profile). Right side: example of baseline and composite electricity load curve for a generic week.

Over the whole year, the maximum peak load increases from 59.6 GW to 64.1 GW (+ 7.5%). With reference to Table 2, the average daily additional load required for electric cooking per household is 1.335 kWh∙household-1∙day-1 (11393.5 kWh∙year-1 in total). By assuming different values for the average efficiencies of induction (0.722) and gas kitchens (0.439) [54], and considering the average daily gas consumption for cooking per family assessed in sub-section 4.1, it is possible to compare the useful cooking energy demand obtained from the model with that from the current cooking sector: these two values, actually supposed to be equal, differs here of about 1.5%, revealing a good degree of accuracy of the proposed approach. Given the electricity load curves (one for the baseline case and for one the new electricity demand), the application the energy optimisation model (sub-section 4.2) returns the cost-optimal electricity production and dispatch strategies for the BAU and the SEN scenarios. Figure 5 reports an example of the energy model results for three days, showing the energy produced by each technology (upper sub-plot), the change in electricity production between the new and the baseline load profiles (middle sub-plot), and the change in Total Primary Energy Supply (TPES, lower sub-plot). Values of TPES are calculated based on the IEA conventions [61], and they also include the avoided natural gas consumption of households’ kitchens.

ACCEPTED MANUSCRIPT In the BAU scenario (Figure 5, left side) the additional cooking load is mainly covered by CCGT plants, and in minor part by other fossil-fuel plants, whereas ramp constraints do not allow CCGT plants to instantaneously cover the whole peak. VRES penetration is too limited to contribute to the new electric load, even in periods of lower baseline load, and PHS storage is rarely employed. As a result, the avoided primary energy consumption for gas use at the residential level is always more than counterbalance by a corresponding higher increase of primary energy consumption within the power sector.

Table 2. Summary of the data and parameters made to test the useful cooking energy obtained from the model with that obtainable from gas consumption data. Item

Units

Value

Natural gas consumption for cooking

GWh

19017.1

Electricity consumption for cooking (model)

GWh

11393.5

Energy efficiency gas cookers

-

0.439

Energy efficiency induction cookers

-

0.722

Useful cooking energy, natural gas

GWh

8348.5

Useful cooking energy, induction

GWh

8226.1

Err%

%

1.5%

Figure 5. Dispatch strategy after the introduction of the additional induction cooking load, and relative difference in power generation and TPES mixes before and after the introduced shock. (left side: BAU scenario; right side: SEN scenario).

Conversely, in the SEN scenario (Figure 5, right side) the higher penetration of VRES significantly contributes to the baseline load (i.e. with no cooking) and often exceeds it, resulting in a significant use of PHS but also in a non-negligible amount of curtailment. Within this energy system configuration, the

ACCEPTED MANUSCRIPT introduction of the additional load required by induction cooking produces a positive shock in the energy system, especially with regards to VRES, which experience a reduction of curtailment and an increased contribution to peak load: indeed, as shown in Figure 5, the timing of PV generation is perfectly matched with the mid-day peak associated with lunch cooking, and when the overall load is in its lower range (e.g. weekends, or in low-demand months), PV generation alone can sustain and even exceed the required midday peak load, even though this entails a reduced use of PHS compared to the baseline-load case. Indeed, it can be inferred from Figure 5 (right side) that, for several instants of time, the power that was generated by PHS before introduction of the additional load is subsequently replaced by CCGT generation, as a consequence of the reduced stored-energy availability. Furthermore, when PV generation is accompanied by a consistent wind generation, the two sources do manage to cover the whole mid-day peak even in days of high overall energy demand. In general, evening peaks associated with dinner cooking are still mainly covered by CCGT plants, except when favourable conditions occur (e.g. low overall demand and high VRES excess that can be stored in PHS plants and used during the evening). Accordingly, CCGT plants still experience an increased use because of the introduced shock, though relatively less marked than in BAU

25 20 15

11.4

10.0

10 5 0

Change in TPES [TWh]

Change in electricity production [TWh]

scenario.

25 20 15 10 5

-1.5

0

-5

-5

-10

-10

-15

-15

-20

-20

-25

2.0

avoided natural gas imports geothermal solar photovoltaic wind hydro oil & derivatives natural gas coal total

-25

BAU

SEN

BAU

SEN

Figure 6. Change in the annual electricity production (left-side) and TPES (right-side), by source, as a result of the introduction of induction kitchens, for both scenarios.

The overall yearly changes in electricity production and Total Primary Energy Supply (TPES) by technology are reported in Figure 6. In the BAU scenario, the change in electricity production (about 11.4 TWh∙y-1) is almost entirely provided by CCGT technology, followed by minor contributions by other fossil-fuel technologies, while there are no power plants reducing their production for cost-optimisation reasons. Despite the avoided natural gas consumption by households’ kitchens, an overall increase in primary energy consumption is expected (about 2 TWh∙y-1). The SEN scenario also results in a strong increase in the use of

ACCEPTED MANUSCRIPT CCGT technology, but also in a slight increase in PV generation – due to the reduced curtailment and/or storage of excess production during the mid-day solar peak – and in a corresponding reduction in the use of PHS. This ultimately results in an expected decrease in primary energy consumption (about -1.5 TWh∙y-1). Results of the energy optimization model provide the basis for the application of the Input-Output model, which finally allows to assess the economy-wide impact of the analysed intervention, here quantified in terms of changes in value-added generation and CO2 emissions. Global results are represented in Figure 7, where the two top graphs focus on Italy, while the two bottom graphs refer to the overall world impacts. The following comments can be made: 

Compared to the BAU scenario, the implementation of inductors kitchens in the SEN scenario is characterised by relatively lower economic and environmental impacts: this confirms the positive synergy between the high VRES penetration and the electrification of final uses.



For both the BAU and the SEN, the environmental and economic impacts due to the implementation of induction kitchens (Electricity category, blue bars) exceed the benefits due to the avoided natural gas consumption in traditional kitchens (Natural Gas category, green bars); this result is even more relevant if manufacturing of inductor cookers is considered (Cookers category, yellow bars). This reflects the fact that natural gas supply chain is significantly shorter and less capital-intensive compared to the electricity one.



The overall economic impact in Italy results in an additional 2.1 B€ (0.10% of Italian 2011 GDP) in the BAU scenario and 1.7 B€∙y-1 (0.08% of Italian 2011 GDP) in the SEN scenario. As explained in sub-section 4.3, these values represent the revenues side of the GDP: they can be interpreted as the increase in factors of production required to implement the analysed policy intervention.



The overall CO2 emissions in Italy increase after the induction kitchens implementation for both the scenarios: 2.07 Mton∙y-1 for BAU and 0.88 Mton∙y-1 for the SEN. This again reflects the higher complexity of the electricity supply chain compared to the natural gas one. Indeed, to obtain the same useful effect (i.e. cooking meals), the production of electricity invokes for more intermediate conversion processes compared to the direct burning of natural gas.



In the SEN scenario, the overall increase in CO2 emissions is not in contradiction with the result obtained by the energy model (Figure 6, right side), where a decrease in primary non-renewable energy is registered. In fact, this can be explained by considering the comprehensiveness of the Input-Output model compared to the energy model: while the latter accounts for the emissions related to the energy sector only, the former includes the impact related to all other sectors.



The analysed intervention always causes non-negligible increases in economic expenditures and CO2 emissions abroad, for both the scenarios: this is motivated by the fact that Italy depends on foreign supply chains for its manufacturing and energy-related activities. However, it can be inferred that a higher penetration of VRES, reflected by the SEN scenario, enables to reduce such economic dependency by about 124 M€∙y-1, and the related CO2 emissions leakage by 137 Mton∙y-1.

4000 3000

2112

2000

ΔCO2 [kton/y]

ΔGDP [M€/y]

ACCEPTED MANUSCRIPT

1664

1000

-448

4000 2000

2068

882

-1186

-4000

-2000

-6000 BAU

SEN

SEN-BAU

4000 3000 627 2112

503 1664

0

-2000

SEN

8000 4000 2000

SEN-BAU Rest of World Italy

6000

0 -448 -124

-1000

BAU

ΔCO2 [kton/y]

ΔGDP [M€/y]

6000

-2000

-1000

1000

Cookers Electricity Natural Gas Italy

0

0

2000

8000

697 2068

560 882

-2000

-1186 -137

-4000 -6000

BAU

SEN

SEN-BAU

BAU

SEN

SEN-BAU

Figure 7. Expected national changes in yearly GDP (left side) and CO2 emissions (right side). The top charts focus on internal changes of the Italian economy, whilst the bottom charts show also interactions with the rest of the world economies. The values are reported for each scenario, and the relative difference between the two scenarios is also displayed.

The overall results of the Input-Output model are disaggregated by the national sectors of the Italian economy in Figure 8. The major contributions to value added generation come from the Services, Transport and storage, Electricity and Manufacturing sectors. The highest contribution to the GDP increase comes from the Manufacturing sector (+691 M€ in the BAU and +641 M€ in the SEN), in which the contribution related to the induction cookers production plays a major role. An overall reduction in value-added generation comes from the Gaseous fuel supply sector, due to the reduction in natural gas requirements by households. Households sector represent the final consumers: therefore, it is neutral regarding the change in national value added. CO2 emissions are relevant only for the Households sector (-3.71 Mton in both scenarios) and in the Electricity supply sector (+4.98 in BAU and +3.90 in SEN). Manufacturing of cooking devices only represents a minor contribution (slightly more than +0.4 Mton in both scenarios). The remaining CO2 emissions from all the other sectors (again around +0.4 Mton in both scenarios) represent the so-called spillover effects, that is, the indirect contributions of the background infrastructures that support the country’s activities.

ACCEPTED MANUSCRIPT Business As Usual (BAU) Households JtU - Services H - Transportation and storage G - Waste management F - Construction E - Steam and water supply D2 - Gaseous fuels supply D1 - Electricity supply C - Manufacturing B - Mining and quarrying A - Agriculture, forestry, fishing

National Energy Strategy (SEN)

0.0

0.0 430.4

338.1

-133.6 -98.9

239.2

9.6

7.3

-2.3

34.6

28.0

-6.5

13.2

9.2

-4.0

-113.7

-146.4

-32.7

515.4

405.2

691.2 58.8

49.7

-9.1 -0.3

1.0

ΔGDP [M€/y]

-600 -400 -200 0 200 400 600 800 1000 -200-100 0 100

ΔGDP [M€/y]

-3710.2 39.0

30.6

308.1

249.3

0.8

0.6

SEN-BAU

Natural Gas Electricity Cookers Total

0.0 -8.5 -58.8 -0.2

6.3

5.1

-1.2

27.4

17.2

-10.3

-7.7

-9.9

-2.2

4983.1

0

-1086.6

3896.4

408.1

-6000 -4000 -2000

-110.2 -50.6

640.7

1.3

-600 -400 -200 0 200 400 600 800 1000 Households -3710.2 JtU - Services H - Transportation and storage G - Waste management F - Construction E - Steam and water supply D2 - Gaseous fuels supply D1 - Electricity supply C - Manufacturing B - Mining and quarrying A - Agriculture, forestry, fishing

0.0

564.0

396.0

-12.1

12.8

6.6

-6.2

0.4

0.3

-0.1

2000 4000 6000 -6000 -4000 -2000

ΔCO2 [kton/y]

0

2000 4000 6000 -2000

ΔCO2 [kton/y]

0

SEN-BAU

Figure 8. Results of the Input-Output model disaggregated by national sectors. Change in GDP is represented in the upper side, while change in CO2 emissions in the bottom side. The values are reported for each scenario, and the relative difference between the two scenarios is also displayed

6. Conclusions This study conceptualised a multi-layer modelling methodology that allows to assess the economy-wide economic and environmental impacts of heat-electricity integration strategies while keeping a highly detailed representation of the energy sector operation, based on the soft-link between three layers of modelling: (i) an electricity load curves estimation model, (ii) an energy optimisation model and (iii) an Input-Output model. The proposed methodology has been used to assess the impact due to the deep electrification of residential kitchens in Italy, relying on: a stochastic model for determining the change in electricity load profile, the Calliope energy optimisation model and the Exiobase v.3 Input-Output database. From a methodological standpoint, this research demonstrates the need for and the advantages of a comprehensive integrated multi-layer modelling approach. In fact, the latter emerges as a fundamental tool to support informed and non-biased policies, which can assess the impact of decarbonisation strategies within the complexity of an integrated smart energy system interacting with the other productive sectors of the economy. This is supported by the obtained quantitative results, from which the following concluding remarks can be derived and summarised:

ACCEPTED MANUSCRIPT 

Results of the analysed case study reveal that, in both the BAU and the SEN scenarios, the substitution of traditional gas kitchens with induction kitchens causes an overall increase in valueadded generation and CO2 emissions.



Such comprehensive multi-layer results expand and complement those achievable from the adoption of the energy model alone; the positive net effect on the energy system arising from the cooking electrification strategy in the SEN scenario is offset, in an economy-wide perspective, by feedbacks on the other productive sectors, as detected by the third layer of modelling (Input-Output model).



In general, the analysis confirms the subsistence of a synergic effect between electrification strategies and higher penetration of VRES in the energy system: overall emission entailed by the intervention are reduced from 2.07 Mton∙y-1 to 0.88 Mton∙y-1 as the current energy system configuration is changed to the one prospected by the SEN, and a similar trend is reported for imports dependency, as a consequence of the lower need for additional fossil-fuel plants operation.

Based on the approach here introduced, some further developments may be considered to push forward the analysis towards the assessment of potentially higher shares of VRES in heat-electricity integration strategies and to provide further guidance to policy makers: 

The adoption of increasing spatial and technical detail in the characterisation of the electricity network (e.g. multi-node electricity model, accounting for transmission constraints and losses), which would allow modelling scenarios with even higher shares of renewables. It is worth noting, however, that accounting for network constraints would not significantly change the results obtained for the considered scenarios, as electricity constraints and losses would more than counterbalance those related to the gas network.



The inclusion in the analysis of other decarbonisation options of the cooking sector, such as powerto-gas (P2G), which may be more suited for a country characterised by an already existing massive gas transmission network, while still allowing for sector coupling and increased use of renewables.

Appendix 1 – Parameters and assumptions of the cooking load model Table 3 reports all the parameters required by the stochastic model to compute aggregate cooking loads. As already mentioned in sub-section 4.1, parameters such as Meal_behaviourji and %out-of-home mealsjik have been gathered from FIPE and Confcommercio [50]. Cooking cycles have been instead modelled based on representative cycles for typical Italian meals, for different meal types (breakfast, lunch and dinner), assessed based on the report by CENSIS and Coldiretti [51] as well as on the authors’ personal knowledge of the Italian cooking habits. The association between cooking cycles and power levels required by induction kitchens has been realised thanks to the information available in the datasheets of several induction kitchens manufacturers, and in some cases counter-checked by means of the Cook-StePS software [62]. All parameters related to likely time frames for each meal type and to ranges of variability to be employed within the stochastic randomisation process have been estimated based on the abovementioned sources.

ACCEPTED MANUSCRIPT The algorithm implemented in the model, briefly summarised by Figure 3 and freely available as open-source code at SESAM Polimi – Italian cooking sector paper [46], revolves around few key concepts: 

Cooking events of a certain type (i.e. switch-on of induction kitchen for lunch) can occur only within pre-defined time frames (subject to stochastic randomisation), and only if the control about the probability of the meal being consumed out of home is passed;



Each cooking event of a given type is constrained to have a minimum duration (t_minjic), whereas the maximum duration is randomised based on cooking-cycle shape and the allowed ranges of variability;



Cooking events of a given type are repeated until reaching a pre-defined (yet subject to randomisation) total time of use during the day (Tot_usejic). For instance, the use of induction kitchen for breakfast (typically involving simple boiling of water for tea or heating up milk) is assumed having a minimum switch-on time of 3 minutes and a maximum time of use throughout the day of 15 minutes ± 10% variability (including also possible tea/milk preparation in the afternoon).

Table 3. Exogenous parameters required by the bottom-up model for the stochastic generation of induction cooking load profiles. Variable name

Description

User_typej

Total number of users for a given user type (e.g. “small family”, “large family”)

Meal_behaviourji

Number of meal types, for each User type, for which a different behavior can be identified

%out-of-home mealsjik

Shares of users that consume a given meal out-of-home for k different frequency classes

Cooking taskjic

Name of the c-th cooking task associated with the j-th user type and the i-th meal behaviour

Cooking cyclejic

Power absorbed by the induction and duration of each segment of the c-th task’s cycle

δcycle,jic [%]

Percentage random variability applied to the duration of the segments composing cyclejic

δP_induction,jic [%]

Percentage random variability applied to Pjik, conceived for thermal appliances

Tot_usejic [min]

Daily average total time spent for the c-th cooking task

t_minjic [min]

Minimum time that the induction is kept on after a switch-on event of the c-th cooking task

δt_min,jic [%]

Percentage random variability applied to t_minjic

use_framesjic

Time frames in which a random switch-on of Cooking taskjic can occur

δframes,jic [%]

Percentage random variability applied to use_framesjic

Abbreviations BAU

Business as Usual

CCGT Combined Cycle Gas Turbine GHG

Greenhouse Gases

IAM

Integrated Assessment Model

IOA

Input-Output Analysis

PHS

Pumped Hydro Storage

ACCEPTED MANUSCRIPT PV

Photovoltaic

SEN

National Energy Strategy

TPES Total Primary Energy Supply VRES Variable Renewable Energy Sources

References [1]

UNFCCC. Paris Agreement. 2015.

[2]

Lund H, Andersen AN, Østergaard PA, Mathiesen BV, Connolly D. From electricity smart grids to smart energy systems - A market operation based approach and understanding. Energy 2012;42:96– 102. doi:10.1016/j.energy.2012.04.003.

[3]

Mathiesen B V., Lund H, Connolly D, Wenzel H, Ostergaard PA, Möller B, et al. Smart Energy Systems for coherent 100% renewable energy and transport solutions. Appl Energy 2015. doi:10.1016/j.apenergy.2015.01.075.

[4]

Bloess A, Schill WP, Zerrahn A. Power-to-heat for renewable energy integration: A review of technologies, modeling approaches, and flexibility potentials. Appl Energy 2018;212:1611–26. doi:10.1016/j.apenergy.2017.12.073.

[5]

European Commission. An EU strategy on Heating and Cooling 2016. vol. 53. 2016.

[6]

Eurostat. Energy consumption in households 2017.

[7]

Brown TW, Bischof-niemz T, Blok K, Breyer C, Lund H, Mathiesen B V. Response to ‘ Burden of proof : A comprehensive review of the feasibility of 100 % renewable-electricity systems .’ Renew Sustain Energy Rev 2018;92:834–47. doi:10.1016/j.rser.2018.04.113.

[8]

Lund H, Østergaard PA, Connolly D, Ridjan I, Mathiesen BV, Hvelplund F, et al. Energy Storage and Smart Energy Systems. Int J Sustain Energy Plan Manag 2016;11:3–14. doi:10.5278/ijsepm.2016.11.2.

[9]

MISE, MATTM. Strategia Energetica Nazionale (SEN). 2017.

[10]

IEA. World Energy Outlook 2018. 2018.

[11]

Pfenninger S, Keirstead J. Renewables, nuclear, or fossil fuels? Scenarios for Great Britain’s power system considering costs, emissions and energy security. Appl Energy 2015;152:83–93. doi:https://doi.org/10.1016/j.apenergy.2015.04.102.

[12]

Hilpert S, Kaldemeyer C, Krien U, Günther S, Wingenbach C, Plessmann G. The Open Energy Modelling Framework (oemof) - A new approach to facilitate open science in energy system modelling. Energy Strateg Rev 2018;22:16–25. doi:10.1016/j.esr.2018.07.001.

[13]

Kavvadias K, Hidalgo Gonzalez I, Zucker A, Quoilin S. Integrated modelling of future EU power and

ACCEPTED MANUSCRIPT heat systems. JRC Tech Rep 2018. doi:10.2760/860626. [14]

Brown T, Jonas H, Schlachtberger D. PyPSA : Python for Power System Analysis. Arxiv 2017.

[15]

Pfenninger S, Hawkes A, Keirstead J. Energy systems modeling for twenty-first century energy challenges. Renew Sustain Energy Rev 2014;33:74–86. doi:10.1016/j.rser.2014.02.003.

[16]

Clegg S, Mancarella P. Integrated electricity-heat-gas modelling and assessment, with applications to the Great Britain system. Part I: High-resolution spatial and temporal heat demand modelling. Energy 2018:1–11. doi:10.1016/j.energy.2018.02.079.

[17]

Wene CO. Energy-Economy Analysis: Linking the Macroeconomic and Systems Engineering approaches. Energy 1996;21:809–24. doi:10.1016/0360-5442(96)00017-5.

[18]

Fortes P, Pereira R, Pereira A, Seixas J. Integrated technological-economic modeling platform for energy and climate policy analysis. Energy 2014;73:716–30. doi:10.1016/j.energy.2014.06.075.

[19]

Bobmann T, Staffell I. The shape of future electricity demand: Exploring load curves in 2050s Germany and Britain. Energy 2015;90:1317–33. doi:10.1016/j.energy.2015.06.082.

[20]

Calise F, D’Accadia MD, Barletta C, Battaglia V, Pfeifer A, Duic N. Detailed modelling of the deep decarbonisation scenarios with demand response technologies in the heating and cooling sector: A case study for Italy. Energies 2017;10. doi:10.3390/en10101535.

[21]

Reiner Lemoine Institut. demandlib n.d.

[22]

Ringkjøb HK, Haugan PM, Solbrekke IM. A review of modelling tools for energy and electricity systems with large shares of variable renewables. Renew Sustain Energy Rev 2018;96:440–59. doi:10.1016/j.rser.2018.08.002.

[23]

Pfenninger S, Hirth L, Schlecht I, Schmid E, Wiese F, Brown T, et al. Opening the black box of energy modelling : Strategies and lessons learned 2018;19:63–71. doi:10.1016/j.esr.2017.12.002.

[24]

Pfenninger S, DeCarolis J, Hirth L, Quoilin S, Staffell I. The importance of open data and software: Is energy research lagging behind? Energy Policy 2017;101:211–5. doi:https://doi.org/10.1016/j.enpol.2016.11.046.

[25]

Pfenninger S. Energy scientists must show their workings. Nature 2017;542:393.

[26]

Quoilin S, Nijs W, Zucker A. Evaluating flexibility and adequacy in future EU power systems : model coupling and long-term forecasting n.d.

[27]

Jiménez Navarro JP, Kavvadias KC, Quoilin S, Zucker A. The joint effect of centralised cogeneration plants and thermal storage on the efficiency and cost of the power system. Energy 2018;149:535–49. doi:10.1016/j.energy.2018.02.025.

[28]

Pavičević M, Tomić I, Quoilin S, Zucker A, Krajačić G, Pukšec T, et al. Applying the Dispa-SET Model on the Western Balkans Power Systems. 12th Conf Sustain Dev Energy, Water Environ Syst 2017:53.

ACCEPTED MANUSCRIPT [29]

Prina MG, Cozzini M, Garegnani G, Manzolini G, Moser D, Filippi Oberegger U, et al. Multi-objective optimization algorithm coupled to EnergyPLAN software: The EPLANopt model. Energy 2018;149:213–21. doi:10.1016/j.energy.2018.02.050.

[30]

Pfenninger S. Dealing with multiple decades of hourly wind and PV time series in energy models: A comparison of methods to reduce time resolution and the planning implications of inter-annual variability. Appl Energy 2017;197:1–13. doi:https://doi.org/10.1016/j.apenergy.2017.03.051.

[31]

OpenMod. Open Energy Modelling Initiative wiki - Open Models n.d.

[32]

Miller RE, Blair PD. Input-output analysis: foundations and extensions. Cambridge University Press; 2009.

[33]

Klinge Jacobsen H. Integrating the bottom-up and top-down approach to energy–economy modelling: the case of Denmark. Energy Econ 1998;20:443–61. doi:10.1016/S0140-9883(98)00002-4.

[34]

Bauer N, Baumstark L, Leimbach M. The REMIND-R model: the role of renewables in the low-carbon transformation—first-best vs. second-best worlds. Clim Change 2012;114:145–68. doi:10.1007/s10584-011-0129-2.

[35]

E3MLab/ICCS at National Technical University of Athens. PRIMES - Detailed Model Description. Athens: 2014.

[36]

Gargiulo M, Gallachóir BÓ. Long-term energy models: Principles, characteristics, focus, and limitations. Wiley Interdiscip Rev Energy Environ 2013;2:158–77. doi:10.1002/wene.62.

[37]

Rocco M V, Casalegno A, Colombo E. Modelling road transport technologies in future scenarios: Theoretical comparison and application of Well-to-Wheels and Input-Output analyses. Appl Energy 2018;232:583–97. doi:https://doi.org/10.1016/j.apenergy.2018.09.222.

[38]

Rocco M V, Langè S, Pigoli L, Colombo E, Pellegrini LA. Assessing the energy intensity of alternative chemical and cryogenic natural gas purification processes in LNG production. J Clean Prod 2019;208:827–40. doi:https://doi.org/10.1016/j.jclepro.2018.10.108.

[39]

Rady YY, Rocco M V, Colombo E, Serag-Eldin MA. Soft-linking Bottom-up energy models with Topdown Input-Output models to assess the environmental impact of future energy scenarios. Model Meas Control C 2018;79:103–10. doi:https://doi.org/10.18280/mmc_c.790307.

[40]

Messner S, Schrattenholzer L. MESSAGE–MACRO: linking an energy supply model with a macroeconomic module and solving it iteratively. Energy 2000;25:267–82.

[41]

Kober T, Summerton P, Pollitt H, Chewpreecha U, Ren X, Wills W, et al. Macroeconomic impacts of climate change mitigation in Latin America: A cross-model comparison. Energy Econ 2016;56:625– 36. doi:10.1016/J.ENECO.2016.02.002.

[42]

Heinrichs HU, Schumann D, Vögele S, Biß KH, Shamon H, Markewitz P, et al. Integrated assessment of a phase-out of coal-fired power plants in Germany. Energy 2017;126:285–305. doi:10.1016/J.ENERGY.2017.03.017.

ACCEPTED MANUSCRIPT [43]

Tukker A, Wood R, Giljum S. Relevance of Global Multi Regional Input Output Databases for Global Environmental Policy: Experiences with EXIOBASE 3. J Ind Ecol 2018. doi:10.1111/jiec.12767.

[44]

Merciai S, Schmidt J. Methodology for the Construction of Global Multi-Regional Hybrid Supply and Use Tables for the EXIOBASE v3 Database. J Ind Ecol 2018. doi:10.1111/jiec.12713.

[45]

ISTAT. ISTAT - Censimento Popolazione 2011.

[46]

Lombardi F, Rocco M V, Colombo E. SESAM Polimi - Italian cooking sector paper 2018.

[47]

Terna. Terna - Dati Statistici. 2015.

[48]

ENTSO-E. ENTSO-E - Historical Data 2015.

[49]

Data OPS. Data Platform - Time Series 2015.

[50]

FIPE, Confcommercio, EBNT. Osservatorio sui consumi fuori casa. 2016.

[51]

CENSIS, Coldiretti. Primo rapporto sulle abitudini alimentari degli italiani. 2010.

[52]

Lombardi F. Cook‐STePS: a tool to expand lab tests information by means of thermodynamic simulations. ETHOS 2018 Abstr., Seattle: 2018.

[53]

International Energy Agency. IEA Statistics - Italy n.d.

[54]

US Department of Energy. Energy Conservation Program: Test Procedures for Conventional Cooking Products. vol. 79. 2014.

[55]

Prina MG, Manzolini G, Moser D, Sparber W. Renewable Energy High Penetration Scenarios Using Multi-Nodes Approach: Analysis for the Italian Case. 33rd Eur Photovolt Sol Energy Conf Exhib 2017:2164–70. doi:10.4229/EUPVSEC20172017-6EO.2.1.

[56]

Prina MG, Fanali L, Manzolini G, Moser D, Sparber W. Incorporating combined cycle gas turbine flexibility constraints and additional costs into the EPLANopt model: The Italian case study. Energy 2018;160:33–43. doi:10.1016/j.energy.2018.07.007.

[57]

Pfenninger S, Staffell I. Long-term patterns of European PV output using 30 years of validated hourly reanalysis and satellite data. Energy 2016;114:1251–65. doi:10.1016/j.energy.2016.08.060.

[58]

Staffell I, Pfenninger S. Using bias-corrected reanalysis to simulate current and future wind power output. Energy 2016. doi:10.1016/j.energy.2016.08.068.

[59]

RSE. Energia elettrica, anatomia dei costi. 2015.

[60]

Christian B, Griffiths T. Chapter 7: Overfitting. Algorithms to Live By Comput. Sci. Hum. Decis., London: William Collins; 2016, p. 149–68.

[61]

IEA. Energy Statistics–Manual. 2005. doi:10.1787/9789264033986-en.

[62]

Lombardi F, Colombo L, Colombo E. Design and validation of a Cooking Stoves Thermal Performance Simulator ( Cook-STePS ) to simulate water heating procedures in selected conditions. Energy 2017;141:1384–92. doi:10.1016/j.energy.2017.11.045.

ACCEPTED MANUSCRIPT 1

A multi-layer energy modelling methodology to assess the

2

impact of heat-electricity integration strategies: the case of the

3

residential cooking sector in Italy

4

Francesco Lombardia, Matteo Vincenzo Roccoa, Emanuela Colomboa

5

aPolitecnico

6

Corresponding author: Lombardi F, Tel.: +39-02-2399-3866; address: Via Lambruschini 4, 21056 Milan, Italy.

7

E-mail: [email protected]

di Milano, Department of Energy, via Lambruschini 4, Milan, Italy

8 9 10

Highlights:

11



A multi-layer methodology to link energy models and input-output models is proposed.

12



Italian electricity demand due to massive electrification of cooking devices is assessed.

13



Change in the Italian power production mix is assessed based on Calliope energy model.

14



The related economic/environmental implications are assessed with the Exiobase model.

1