High Resolution Generation Expansion Planning Considering Flexibility Needs: The Case of Switzerland in 2030
Elena Raycheva, Jared Garrison, Christian Schaffner, Gabriela Hug
aa r X i v : . [ m a t h . O C ] J a n HIGH RESOLUTION GENERATION EXPANSIONPLANNING CONSIDERING FLEXIBILITY NEEDS:THE CASE OF SWITZERLAND IN 2030
Elena Raycheva * , Jared Garrison , Christian Schaffner , Gabriela Hug EEH - Power Systems Laboratory, ESC - Energy Science Center, FEN - Research Center for Energy Networks,ETH Zurich, Zurich, Switzerland*E-mail: [email protected]
Keywords:
GENERATION EXPANSION PLANNING, UNIT COMMITMENT, RESERVES, RENEWABLESINTEGRATION, NODAL DISPATCH
Abstract
This paper presents a static generation expansion planning formulation in which operational and investment decisions for a widerange of technologies are co-optimized from a centralized perspective. The location, type and quantity of new generation andstorage capacities are provided such that system demand and flexibility requirements are satisfied. Depending on investments innew intermittent renewables (wind, PV), the flexibility requirements are adapted in order to fully capture RES integration costsand ensure normal system operating conditions. To position candidate units, we incorporate DC constraints, nodal demand andproduction of existing generators as well as imports and exports from other interconnected zones. To show the capabilities of theformulation, high-temporal resolution simulations are conducted on a 162-bus system consisting of the full Swiss transmissiongrid and aggregated neighboring countries.
The integration of large RES capacities has a significant impacton power systems planning as it increases the need for opera-tional flexibility from existing and future units. Consequently,the change in system reserve requirements in response to theexpected growth of RES has to be accounted for. The mainobjective of this work is to present a formulation of the Gener-ation Expansion Planning (GEP) problem which accounts forboth present and future flexibility needs and demonstrate itsfunctionality on a real-size power system.The goal of GEP is to determine the optimal investmentsin new generation and storage technologies over a certainplanning horizon, in order to meet load growth and replacedecommissioned units. A detailed review of GEP in the contextof increasing integration of RES is presented in [1]. Planninggeneration expansion in power systems with large shares ofRES requires modeling the detailed operational constraints ofboth existing and candidate technologies providing flexibilitysuch as storages, hydro and thermal generators. Furthermore,reserve constraints have to be included to fully capture the costsof integrating intermittent generation and ensure normal systemoperating conditions. While [2]-[5] recognize this and handlesome of the constraints, they focus primarily on thermal unitsand do not consider hydro or storages as sources of operationalflexibility. In contrast, the present work includes different typesof flexibility providers.Further flexibility can be provided through imports andexports from other interconnected zones. This is currentlyconsidered to be the most convenient and cheapest way toincrease flexibility in regions with reliable grid connectivity [6]. Thereby, it is important to model market-based tie lineflow constraints as opposed to the full cross-border line limitsto better reflect the realistic ability to import/export. Modelingthe grid within the considered zone including its connectionto neighboring countries, allows to position candidate units atsystem nodes of interest and determine favorable locations toalleviate, for example, grid congestions. In [7] and [8], DCconstraints are included in the investment model, using sim-plified transmission systems as test cases. In contrast, our workincludes a study of the capability of a real-size power systemto evolve and cope with the projected increase in intermittentRES capacities. We address uncertainties related to renewableproduction by modeling the reserve provision of both existingand candidate units and capture the increased needs and costsin terms of tertiary system reserve requirements in the opti-mization problem. This is similar to [9], however, in the presentwork we use nodal dispatch, market-based limits on the cross-border tie lines as well as a unit commitment formulation forthe operation of conventional thermal generators, which hasbeen shown to have a significant impact on investment deci-sions [8].An additional novelty of the work lies in the high tempo-ral and spacial resolution of the conducted simulations coupledwith detailed modeling of flexibility provision, spanning 1)imports/exports from other zones, 2) operation and 3) reserves.The case study in this paper presents two possible scenarios,namely with and without fixed RES target, for the devel-opment of the Swiss generation portfolio in the context ofnuclear phase-out and desired increase in RES production. Asgeneration in Switzerland, but also in many other regions, isdominated by hydro capacities, a high temporal resolution for oth the validation and generation expansion is needed. Whilein [2]-[5], [7]-[8] a few representative days or weeks are used,in the present work every other day of the year is simulatedwith hourly resolution. We validate the short-term operationformulation to ensure agreement with historical data since anydiscrepancies could impact future investment decisions. Theremainder of the paper is organized as follows: the problemformulation is in Section 2. The case study and results are inSection 3 while conclusions are drawn in Section 4. In the following problem formulation lowercase letters are usedto denote variables and uppercase letters denote parameters.The objective of the generation expansion planning problemis to minimize the sum of the production and investment costsof all existing and candidate generation and storage technolo-gies over the planning horizon T , which in the present work isfixed to a single year: min X j ∈ J X t ∈ T ( C prodj p j,t + C suj v j,t ) | {z } i) + X s ∈ S X t ∈ T C prods p diss,t | {z } ii) + (1) X r ∈ R X t ∈ T C prodr p r,t | {z } iii) + X n ∈ N X t ∈ T C ls ls n,t | {z } iv) + X c ∈ C u invc I c | {z } v) where i) - iii) are the production costs of the set of thermalgenerators J , energy storage systems S and non-dispatchablerenewable generators R , iv) refers to the load shedding costsat the N system nodes and v) are the investment costs asso-ciated with building new units from the set of candidate units C . All production costs are assumed to be linear functions ofthe power generated by the given thermal unit, storage sys-tem or renewable generator and the associated operational costparameter C prod . The start-up costs of all thermal generatorsare expressed as linear functions of the cost parameter C suj and the startup binary variable v j,t and are independent ofthe time since last shut-down. For energy storages, only theoperational cost associated with purchasing electricity duringcharging (pumping) are included. The load shedding at any sys-tem node n is the product of the load shedding cost parameter C ls and the load shedding variable ls n,t . In the investment costformulation, u invc denotes the investment decision for each can-didate unit c . The investment cost I c is annualized to accountfor differences in lifetime. Expression (1) is subject to foursets of constraints related to: 1) short-term operation, 2) invest-ments, 3) reserve provision, and 4) transmission system, all ofwhich are described in the following. Short-term operation is modeled by incorporating both the pro-duction as well as the reserve provision capabilities of thermalgeneration, storage and non-dispatchable RES technologies.Table 1 shows which technology types can contribute towardssecondary (SCR) and tertiary (TCR) control reserve. Primary reserve is not explicitly modeled as it constitutes less than 10%of the total hourly reserve quantity and in many western EUcountries it does not have to be procured locally [10].Table 1 Reserve provision per technology
Technology SCR ↑↓ TCR ↑↓ Thermal: Nuclear/Gas/Coal, etc. ✓ ✓
Storage: Pumped Hydro/Dam ✓ ✓
Storage: Battery ✓ ✗
RES: PV/Wind/Run-of-River ✗ ✗
The Unit Commit-ment (UC) constraints of thermal generators are based on thetight and compact formulation in [11] and use three binary vari-ables u j,t , v j,t , w j,t , respectively for the on/off status, start upand shut down and one continuous variable p minj,t for the poweroutput above minimum by each unit j in each time period t [11]. The downward generation constraint is: ≤ p minj,t − ( r SCR ↓ j,t + r TCR ↓ j,t ) , ∀ j, t (2)where r SCR ↓ j,t and r TCR ↓ j,t denote the variables for contribu-tion of each generator towards downward SCR and TCR. Theupward generation constraints are given by: p minj,t + ( r SCR ↑ j,t + r TCR ↑ j,t ) ≤ ( P maxj − P minj ) u j,t − (3) ( P maxj − SU j ) v j,t , ∀ t, ∀ j ∈ M utj = 1 p minj,t + ( r SCR ↑ j,t + r TCR ↑ j,t ) ≤ ( P maxj − P minj ) u j,t − (4) ( P maxj − SD j ) w j,t +1 , ∀ t, ∀ j ∈ M utj = 1 where P max/minj refers to the maximum/minimum power out-put of the conventional generator j and SU j /SD j are itsstart-up/shut-down capabilities. In case the min. uptime of thegenerator, M utj , is two hours or more, a tighter formulation is: p minj,t + ( r SCR ↑ j,t + r TCR ↑ j,t ) ≤ ( P maxj − P minj ) u j,t − (5) ( P maxj − SU j ) v j,t − ( P maxj − SD j ) w j,t +1 , ∀ t, ∀ j ∈ M utj ≥ We further use the ramping and min. up/down time constraintsfrom [11]. Due to space constraints we omit their formulationhere. The logical relation between the different statuses and thetotal power generated are given by: u j,t − − u j,t + v j,t − w j,t = 0 , ∀ t, ∀ j (6) p j,t = P minj u j,t + p minj,t , ∀ t, ∀ j (7)Planned maintenance is modeled using: u j,t ≤ S j,t , ∀ j ∈ J maint , ∀ t (8)where S j,t is the time series indicating the unit’s availabilitythroughout the simulation horizon. The bounds of all previ-ously defined variables are: p minj,t /r SCR/TCR ↑↓ j,t ≥ , u j,t /v j,t /w j,t ∈ [0 , , ∀ t, ∀ j (9) The operational constraints ofeach storage unit are modeled with three continuous vari-ables: p diss,t and p chs,t are used for the discharge (turbine) andcharge (pump) power and are limited by the maximum dis-charge/charge power P max,dis/chs,t . The variable energy level s,t is limited by the energy rating (reservoir energy storagelevel) E maxs and the final storage level E s,T is set equal tothe initial value E s, , i.e. the energy level at the beginningand the end of the year are set equal. It is assumed that eachstorage system can turn on and produce/consume at maximumdischarge/charge power instantaneously: ≤ p dis/chs,t ≤ P max,dis/chs , ∀ t, ∀ s, (10) E mins ≤ e s,t ≤ E maxs , ∀ t, ∀ s and E s,T = E s,T , ∀ s (11) e s,t = e s,t − + η chs p chs,t − p diss,t η diss | {z } ∀ t, ∀ s + ξ s,t |{z} ∀ t, ∀ s ∈ S hyd (12) e s,t ≥ , ∀ t, ∀ s (13)Eq. (12) describes the energy content of each storage unit ineach hour, taking into account the charging/discharging effi-ciencies η chb and η diss . For hydro power plants, the hourlyinflows are included in the term ξ s,t . The upward and down-ward reserve constraints are: r SCR ↑ s,t + r TCR ↑ s,t ≤ P max,diss − p diss,t + p chs,t , ∀ t, ∀ s ∈ S (14) r SCR ↓ s,t + r TCR ↓ s,t ≤ P max,chs − p chs,t + p diss,t , ∀ t, ∀ s / ∈ S dam (15) r SCR/TCR ↑↓ s,t ≥ , ∀ t, ∀ s (16)where for batteries the variable contribution towards tertiaryreserve r TCR ↑↓ s,t is set to zero. Constraints (10)-(14) are alsovalid for hydro dams without pumping capabilities with p chs,t forced to zero. Constraint (14) allows all storage types to pro-vide upward reserve even if they are not producing. Similarly,pumped hydro and batteries can provide downward reservewhile staying idle. This assumption is valid as storage units areconsidered to be infinitely flexible. To ensure that hydro damsdo not provide downward reserve when not producing, we add: ≤ p diss,t − ( r SCR ↓ s,t + r TCR ↓ s,t ) ≤ P max,diss , ∀ s ∈ S dam , ∀ t (17) Production from solar, windand run-of-river power plants is modeled via exogenouslydetermined capacity factor profiles, CF r,t multiplied by theunit’s maximum installed power P maxr . We further allow forcurtailment of renewable power, i.e.: ≤ p r,t ≤ CF r,t P maxr , ∀ t, ∀ r (18) For thermal generators, the investment decision variable u invc from (1) is binary which corresponds to investments in discreteunits. To only dispatch units that have been built, the invest-ment and operational decisions for candidate units are linked: u c,t ≤ u invc , u invc ∈ [0 , , ∀ t, ∀ c ∈ C thermal (19)where u c,t is the binary variable for the on/off status of eachthermal candidate unit in each time step. Similarly, for storages: ≤ p dis/chc,t ≤ u invc P max,dis/chc , u invc ∈ [0 , , ∀ t, ∀ c ∈ C storage (20) To only allow reserve provision by units that are built, we usethe same constraints as for already existing units, but mul-tiply P max,dis/chs in (14)-(15) by u invc . For non-dispatchableRES candidate generators, the investment decision variable u invc is continuous and corresponds to the built capacity at thecandidate location with capacity factor CF c,t and maximumallowable investment capacity P inv,max : ≤ p c,t ≤ u invc CF c,t , ≤ u invc ≤ P inv,maxc , ∀ t, ∀ c ∈ C RES , (21) The formulation of the reserve constraints in Section 2.1 allowsfor non-symmetric reserve provision by each generator/storageunit which is consistent with efforts to reduce market barriersfor smaller bidders who might be unable to offer symmetricalpower bids [10]. The reserves provided by the units have to sat-isfy the system-wide demand for up/down balancing capacityin each time period: X j ∈ J r TCR ↑ j,t + X s ∈ S hydro r TCR ↑ s,t ≥ T CR ↑ ,syst + r TCR ↑ ,RES , ∀ t (22) X j ∈ J r TCR ↓ j,t + X s ∈ S hydro r TCR ↓ s,t ≥ T CR ↓ ,syst + r TCR ↓ ,RES , ∀ t (23)where T CR ↑ ,syst is the upward tertiary system reserve quantityrequired by the Transmission System Operator (TSO) at timestep t . Depending on the investments in wind and solar PVcapacities, an additional tertiary reserve quantity r TCR ↑↓ ,RES is added to ensure that there is enough system flexibility tocompensate uncertainties in RES production: r TCR ↑↓ ,RES = A ↑↓ wind X c ∈ C RESwind u invc + A ↑↓ pv X c ∈ C RESpv u invc (24)where A wind/pv is an empirically derived coefficient calcu-lated following the methodology in [9] where short-term windand PV forecast methods were used to quantify the additionalreserves needed. The constraints for provision of secondaryreserve are identical to (22)-(23) without the additional termsfrom (24). Similar to [12], this formulation assumes that thevariability in RES generation is accounted for in the tertiaryreserve requirement. The following equation models the active power balance ateach bus node n ∈ N where P Dn,t is the nodal demand, ls n,t refers to the load shedding variable, and the remaining termscorrespond to the power output of each generator or storage: p n,t = P Dn,t − ls n,t + X s ∈ S n,t p chs,t − (25) X j ∈ J n,t p j,t − X s ∈ S n,t p diss,t − X r ∈ R n,t p r,t , ∀ n, ∀ t The nodal active power p n,t is the sum of the active power flowsof all lines l ∈ L connected to n as given in: p n,t = X i ∈ l ( n,i ) p l ( n,i ) ,t , ∀ t, ∀ n (26) nd the active power flow p l of a single line is: p l ( n,i ) ,t = B l ( δ n,t − δ i,t ) , ∀ t (27) − P maxl ≤ p l ( n,i ) ,t ≤ P maxl , ∀ t, ∀ l ( n, i ) , (28)where B l is the admittance, δ n , δ i are the voltage angles at thestart and end nodes and P maxl is the thermal limit of the line.Load shedding is allowed at each demand bus: ≤ ls n,t ≤ P Dn,t , ∀ t, ∀ n (29) In this section, the proposed formulation is applied to an inter-connected system consisting of the detailed Swiss transmissionnetwork with aggregated neighboring countries. This modelin total comprises 263 transmission lines, 162 substations, 21transformers and 355 existing generators with total installedcapacity of 460 GW in 2015. As Switzerland’s generationportfolio is heavily dominated by hydro capacities, capturingtheir operational behavior is salient to any model attemptingto replicate historical or predict future production. To speedup computations, while maintaining very high temporal reso-lution (necessary due to short-term fluctuations in river flows,wind and solar generation) and chronological accuracy (nec-essary due to the presence of seasonal storages), every otherday of the year is simulated with hourly resolution. Thus, thechange in demand behavior between weekdays and weekendsduring each week is always captured.Fig.1 shows how hydro storage levels are approximated forthe days which are not simulated. This form of compression isonly used for pumped and dam hydro power plants and not forbattery storages as it is assumed the latter operate on shortercycles (less than a day). Our approach relies on the assump-tion of day-to-day similarity in operation of both pump anddam power plants. This is valid for dams as they operate ona seasonal cycle as well as for pumped power plants which,depending on their reservoir capacity, operate on a daily toweekly cycle. By adapting (11)-(12), the pumping/turbiningacross two days is aggregated into the time during which thestorage charges/discharges in a single day, which means thatthe modeled fluctuations in storage level would have double theamplitude. This doubling is not relevant for seasonal storages,but is relevant for those that operate on a daily cycle. There-fore, the initial/minimum/maximum reservoir levels of dailypumped storages are doubled.To establish the investments and UC schedule, the MILPformulation is implemented in Pyomo [13] and solved withGurobi [14]. The considered investment costs span the entireyear instead of only every second day, therefore we double theoperating costs in (1). As it is possible to get non-unique solu-tions for the hourly operation of hydro storages, stemming fromthe aggregated modeling of the surrounding countries and sim-plified production costs, we fix the investment and binary UCdecisions and re-solve the linear dispatch problem while alsoincluding a negligible price incentive for keeping more waterin the storages as a security measure. In this way, we are able to choose a specific storage curve out of the ones that all leadto the same objective function value. days ( hours) . . .. . . days ( hours)(11) → E mins ≤ e s,t ≤ E maxs , ∀ s ∈ S pump,dayily , ∀ t (12) → e s,t = e s,t − + 2 η chs p chs,t − p diss,t η diss + 2 ξ s,t , ∀ s ∈ S hyd Fig. 1. Days compression for simulation speed-up
A validation for the year 2015 is performed. The generatorsin the surrounding countries are aggregated to one unit pertechnology type. The total capacity per country in 2015 andthe operational constraints used are indicated in Table 2. Thehourly load data and reserve requirements for CH are takenfrom [15]. Load data for the neighbors are from [16]. The cross-border flows between the surrounding and all other countrieswhich are not currently modeled (e.g. DE-DK, etc.) are fixedto the 2015 values from [16]. Solar irradiation and wind timeseries are from [9]. The initial/final storage levels for 2015 andhydro inflows are from [17].Table 2 Generation capacities per country (2015)
Country Detail Gens. Cap. [GW] Constr.
Austria (AT) ✗ * (10)-(13)(18)Germany (DE) ✗
12 187France (FR) ✗
11 110Italy (IT) ✗
10 122Switz. (CH) ✓
313 19 (2)-(24) * For these conventional units we use linear min/max con-straints for p j,t , i.e. P minj ≤ p j,t ≤ P maxj , without model-ing ramping and minimum up/down times.light – 2015 his / dark – 2015 sim monthTWhRun-of-River Nuclear Storage OthersLoadFig. 2. Monthly production per technology type in CH (2015)Fig. 2 compares the 2015 monthly simulated productionin Switzerland to the historical values from [17]. The simu-lation results show quite good agreement with the historicalproduction. The most notable difference is the generation ofhydro storages in August, October and December. In the sim-ulation, more water is stored in the period April-July and is hen turbined in August. This can also be seen in Fig. 3 whichshows the water level over the year. Water levels peak at theend of September. Again more water is stored in October andNovember and used in December. Historically, the dispatchis more evenly distributed. A potential explanation is that thesimulation has perfect foresight about demand, inflows andrenewable production, which power plant operators do not,therefore they enter long-term contracts to hedge their pro-duction. Storing/producing significantly more in any particularmonth exposes the hydro generators to risks. monthTWh HistoricalSimulatedFig. 3. End-of-month storage levels in CH (2015)As future investments in Switzerland could be influenced bythe export/import with its neighbours, it is also important tocorrectly reproduce cross-border flows. Table 3 shows the netSwiss cross-border exchange in 2015. The model is able to cap-ture the most important trend - Switzerland is a transit countryfor flows from Germany, France and Austria to Italy.Table 3 Net CH cross-border exchange (2015) Net Export (From - To) Hist. [TWh] Sim. [TWh]
AT - CH 6.7 4.9DE - CH 13.1 9.0FR - CH 5.3 4.3CH - IT 25.4 21.0Possible reasons for the deviations in absolute magnitude of netcross-border exchange include: the generators and transmissionsystem of the surrounding countries are modeled in aggregateand the assumed variable operation costs are static relative tothe actual ranges of operating costs. Regarding the grid topol-ogy, all CH cross-border lines go to a single node within theneighboring country which means that in each hour there isa single direction of cross-border exchange, which is not thecase in reality. Overall, the presented validation results reflectthe historical values well and give confidence in the model.
The expansion planning is established for Switzerland for thetarget year 2030. Following a 50-year decommissioning plan,only 36% (1220 MW) of the 2015 installed nuclear capacityin the country will remain operational in 2030. We presenttwo different scenarios: 1) business-as-usual (BaU) and 2)renewable target (RES target). In 2) a production target of 9TWh from non-hydro renewable generators (including exist-ing biomass, PV and wind) is imposed in Switzerland. In both scenarios, 65 candidate units with varying sizes and costparameters from [19] and own calculations, summarized inTable 4, are placed at system nodes of interest. While the costsof biomass reflect on-going waste incineration subsidies whichare expected to continue in the future, we assume no subsidiesfor PV and wind. All planned hydro power and transmissionsystem upgrades in the period 2016-2025 are included. Swissdemand and fuel cost projections for 2030 are from [18] and[20]. Hydro inflows are set to the 2015 values and the produc-tion profiles for PV and wind candidates are from [9].Table 4 Cost parameters of candidate units in CH (2030)
Tech. Invest. Cost[kEUR/MW/a] Var. Cost[EUR/MWh] Cap.[MW] Units
Gas CC 84 85 4200 28Gas SC 54 131.5 600 14Biomass 125 1 240 12Wind 206 2.5 1510 7PV 106 2.1 10000 4The demand, generators and fuel costs in the surroundingcountries are adapted to reflect the 2030 projections from [20].The 2015 wind and solar production profiles of the neighborsare scaled to match the projected 2030 totals from [20] and thecross-border flows with all other countries are fixed to the val-ues for 2015 [16]. Table 5 summarizes the investments madeunder the two considered scenarios. Even without a RES tar-get, all biomass candidate power plants are built. Given theirlow costs and the decreased nuclear production, it is more eco-nomically viable to have new generators produce locally than tosolely import. To satisfy the target, a total of 240 MW biomassand 3254 MW PV is invested in (the remaining 3.36 TWh areproduced by existing generators: 2.1 TWh (biomass), 1.1 TWh(PV) and 0.16 TWh (wind)). As a result of the increased inter-mittent RES generation in the second scenario, the total tertiarysystem reserve requirements in each hour increase by 26 MW(up) and 28 MW (down). Even with this increase, investmentsin new dispatchable units were not needed.Table 5 New investments in CH (2030)
Scen. Techn. Built[MW] Gen.[TWh] + TCR ↑ [MW] + TCR ↓ [MW] BaU Biomass 240 2.0 ✗ ✗
RES BiomassPV 2403254 2.03.64 ✗ ✗ and fuel costs. Since Switzerland is a pricetaker during the majority of the year, the generation costs ofthe conventional units in the surrounding countries have a pro-found impact on Swiss electricity prices.Fig. 4 compares the past (2015) and future (2030) monthlysimulated generation. The largest differences between 2015and 2030 occur during the winter months (Nov-Mar) whenthe reactors in 2015 are producing at high levels and there is able 6 Change in net gen. and av. el. price in CH (2030) Scen. Tot. net gen. [% 2015] Av. el. price [% 2015]
BaU -19% +51%RES -14% +47%light – 2015 / normal - 2030 BaU / striped – 2030 RES monthTWhRoR Nuclear Storage BiomassPV Wind OthersFig. 4. Monthly simulated production per technology in CHless PV generation, and in June. In both future scenarios, theremaining Swiss nuclear reactor Leibstadt is shut down forscheduled refueling in June. As a result, the 2015 productioncan not be reached despite the high solar output and Switzer-land becomes a net importer during this month which used tobe an export month in 2015. In 2030, the majority of electricityimports come from France as opposed to Germany (in 2015).This is due to the complete nuclear phaseout in Germany andthe projected growth of RES in France.
This paper presents a GEP formulation which provides thelocation, type and size of new generators/storages consider-ing system flexibility needs. The multitude of results and theirhigh level of detail (both spatial and temporal) could prove tobe useful to TSOs, policy makers and asset owners/operatorsalike. Future work will focus on including more details inthe modeling of the surrounding countries and hydro powerin Switzerland. Furthermore, we will investigate a coordi-nated approach to investments in new generation capacity onthe transmission and distribution system levels to improve themodeling of PV integration in Switzerland.
The authors would like to thank the Swiss Federal Officeof Energy (SFOE) for supporting the Nexus-e project (Nr.SI/501460). The views and opinions in this paper are those ofthe authors and do not necessarily reflect the position of SFOE.6