Electricity portfolio planning model incorporating renewable energy characteristics

Applied Energy 119 (2014) 278–287

Contents lists available at ScienceDirect

Applied Energy journal homepage: www.elsevier.com/locate/apenergy

Electricity portfolio planning model incorporating renewable energy characteristics Jung-Hua Wu a, Yun-Hsun Huang b,⇑ a b

Department of Resources Engineering, National Cheng Kung University, Tainan 701, Taiwan Industrial Economics and Knowledge Center, Industrial Technology Research Institute, Hsinchu 310, Taiwan

h i g h l i g h t s  Traditional electricity planning is only based on the least-cost principle.  The study incorporates the features of renewable energies into electricity planning.  Results exhibit that renewable energies can hedge against fossil fuel price risk.  Results show that the use of renewable energy contributes greatly to reducing CO2.  Reflecting the intermittency requires LNG-fired units to serve as backup generators.

a r t i c l e

i n f o

Article history: Received 4 January 2013 Received in revised form 25 November 2013 Accepted 2 January 2014 Available online 28 January 2014 Keywords: Electricity planning model Portfolio theory Learning curve Capacity credit Intermittency

a b s t r a c t Traditional electricity planning models pursue minimal costs, yet their design often results in an underestimation of the true benefits of renewable energy. This paper attempts to introduce different complementary approaches to traditional electricity planning model to incorporate various renewable energy characteristics and uses Taiwan’s electricity sector as a case study. The portfolio theory, learning curve theory and the capacity credit are applied in the proposed model to reflect characteristics of renewable energy, such as a hedge against fossil fuel price volatility, significant technological progress, and intermittent generation. Simulation results demonstrate that using renewable energies has the advantage of hedging against the volatile fossil fuel price risk as well as reducing carbon dioxide emissions. Considering the intermittency of renewable energies requires LNG-fired plants to serve as the backup generators. However, wind power can only account for limited share of total installed capacity due to the limited land resources in Taiwan. Therefore, taking intermittency into account only demonstrates a small influence of the reserve margins of the entire power system and the total generation costs. Ó 2014 Elsevier Ltd. All rights reserved.

1. Introduction The impact of conventional fossil fuels on the environment has led to a search for a reliable and plentiful energy source that creates less pollution while still allowing for continued economic growth. Renewable energy comes from natural sustainable resources and emits a low amount of pollution. Energy diversification decreases dependency on imported fuels [1]; consequently, renewable energy development has become an important focus in countries around the world, including Taiwan [2]. Conventionally, electricity planning is based on least-cost principles, in which systems perform in environments of relative cost certainty, relatively slow technological progress, high availability of homogeneous electricity generating technologies and stable ⇑ Corresponding author. Tel.: +886 3 5914745; fax: +886 3 5820095. E-mail address: [email protected] (Y.-H. Huang). 0306-2619/$ - see front matter Ó 2014 Elsevier Ltd. All rights reserved. http://dx.doi.org/10.1016/j.apenergy.2014.01.001

energy prices [3]. However, fossil fuel prices have been fluctuating significantly in recent years. Using a traditional planning model would result in a preference for fossil fuel techniques, thereby overlooking the benefits of renewable energy, including the elimination of price volatility associated with fossil fuels. This approach would be unfavorable for countries that highly depend on imported energy like Taiwan. Recently, rapid technological progress has significantly reduced the cost of renewable energy; consequently, the variable costs and intermittency of this kind of energy lead to increased uncertainty when incorporated into power systems. It is therefore essential to integrate the features of renewable energy into traditional analytical frameworks of electricity planning. The objective of this paper is to construct an electricity portfolio planning model that incorporates the characteristics of renewable energy. Using Taiwan’s electricity sector as a case study, we

J.-H. Wu, Y.-H. Huang / Applied Energy 119 (2014) 278–287

applied portfolio theory, learning curve theory,1 and the capacity credit2 to reflect the characteristics of renewable energy, including a hedge against fluctuation in fossil fuel prices, significant technological progress, and intermittent generation. The proposed model was designed to minimize ‘‘the present value of total generation costs after risk adjustment’’ which considers both the present value of total generation costs and the risk (i.e. variance of total generation costs). The model also factors in the constraints of traditional electricity planning models. We then performed simulation analyses and observed the technology portfolios and total generation costs in various scenarios. Finally, suggestions for future policy-making related to the electricity sector are proposed. The paper is structured into seven sections. The following section begins with a review of relevant literature. Section 3 introduces overview of the current status of Taiwan’s electricity sector. Section 4 outlines the structure of the model. Section 5 contains data sources and adjustments. Section 6 presents the simulation results and the final section provides our conclusions.

2. Literature review Traditional electricity planning models pursue minimal costs, yet their design often results in an underestimation of the true benefits of renewable energy. Researchers in a number of countries, including Japan [3], China [13], Turkey [14], the UK [15], Italy [16], Spain [17], and those of the EU [18,19], have applied portfolio theory to identify the most effective portfolio of power generation technologies. Their results show that traditional models often overlook and are unable to consider the value of renewable energy as a hedge against fossil fuel price risk. In view of this, this paper attempts to combine portfolio theory with traditional electricity planning models to demonstrate the benefits of renewable energy technologies in reducing the generation cost-related risks in the electricity sector. Researchers in Taiwan primarily employed multi-objective (e.g. the MULTEEE model [20]) or single-objective models (e.g. the MARKAL model [21]) to evaluate renewable energy issues. The objective functions of these models are set for either minimum costs or minimum carbon dioxide emissions (or both). In the selection of electricity generating technologies, there is a lack of consideration for cost-related risks; future fuel costs are assumed to be stable, and the costs of power generation are discounted to obtain levelized costs. For countries which are highly dependent on imported fuel, such as Taiwan, this approach to determine technology portfolios leads to a bias favoring fossil fuels. Furthermore, in technical progress, it is generally assumed that the decrease rates of future costs are exogenously given (for example, the costs decrease by a certain percentage each year) or else the future costs are directly given. Hence, cost reduction caused by the endogenization of technological change is not taken into account. The above are all flaws present in previous electricity planning models. In order to compensate for these shortcomings in existing literature, in this study, we incorporated the features of renewable energy as a 1 In energy and climate models, the learning curve has been employed with increasing frequency to account for cost reductions due to technology related learning and for endogenizing technological change [4–12]. In this paper, we used one-way learning curve for renewable energy technologies, in order to take into consideration their greater potential for unit cost reductions and endogenization of technological change. 2 The capacity credit of that intermittent generation (e.g. wind) means how much of the intermittent generation capacity can be relied on statistically to meet peak demand. That also indicates how much fossil fuel plant can be replaced, while maintaining the same degree of system security, in other words an unchanged probability of failure to meet the reliability criteria for power system. In this paper, we used capacity credit for wind power, in order to take into consideration its contribution to meet peak demand in our long-term electricity planning model.

279

hedge against fossil fuel price risk, endogenous technological changes, and the intermittency of power generation. Finally, the model has been applied for the case study of Taiwan’s electricity sector. The model is also applicable for other countries, while the necessary data for model calibration are available. 3. Overview of the current status of Taiwan’s electricity sector Table 1 shows the total amount and proportions of electricity generated by Taiwan’s electricity sector in 2011. The power supply structure indicates that the overwhelming majority (95.1%) of the electricity in Taiwan is generated using thermal and nuclear energy; renewable energy is responsible for a mere 3.7%. Due to the nuclear-free homeland policy, nuclear waste disposal problems, and concern aroused by the Fukushima nuclear disaster, all nuclear power development, with the exception of the Lungmen Nuclear Power Plant (commonly known as the Fourth Nuclear Power Plant), has been suspended. Nevertheless, the demand for electricity in Taiwan will continue to rise under the economy progresses. As an indigenous energy, the correlation between the price volatility of renewable energy and that of fossil fuels is weak. Therefore, increasing the proportion of renewable energy in the power generation structure can spread the volatility risk of imported fuels and reduce emission of greenhouse gases. In addition, renewable energy technologies, due to their greater potential for cost reductions (higher learning rates), are expected to successfully compete with conventional technologies at some time in the future. However, renewable energies are affected by the time of the day, seasons and weather, and this intermittency will surely affect the continuity and stability of the power supply. Thus, the problem of integrating different features of renewable energies into traditional electricity planning model is crucial. This paper attempts to introduce different complementary approaches to traditional electricity planning model to incorporate various renewable energy characteristics. 4. Model description 4.1. Objective function In our proposed model, we set the objective function as minimal present value of total generation costs after risk adjustment. In other words, the objective was to minimize both the present value of total generation costs and the risk of total generation costs. Total generation costs include power production costs (fuel costs and variable operation and maintenance costs) and power capacity costs (fixed operation and maintenance costs and capital investment cost). Where the model refers to technical progress it is embodied technical progress, which means the technical progress displayed by newer and more efficient capital performance. Under these circumstances, the capital structure is no longer homogeneous and comprises different vintages. We therefore divided the power generation units by vintage3 to present embodied technical progress. The development of mathematical models for total generation costs and the risk are outlined in the following. 4.1.1. Present value of total generation costs The present value of total generation costs (PVTGC) in this model can be expressed as follows.

3

Vintage refers to the year in which the power generation unit was installed.

280

J.-H. Wu, Y.-H. Huang / Applied Energy 119 (2014) 278–287

Table 1 Total electricity generation in Taiwan in 2011 (Unit: GWh). Item

Taipower

IPP

Cogeneration

Total

Coal-fired Oil-fired LNG-fired Nuclear Pumped storage hydro Renewable energy Conventional hydro power Others Total

67896.8 7307.7 45149.0 42129.2 2902.1

23225.5 – 18255.0 – –

32351.8 3371.6 157.9 – –

123474.1 (49.0%) 10679.3 (4.2%) 63561.9 (25.2%) 42129.2 (16.7%) 2902.1 (1.2%)

3883.2 812.1 170080.1 (67.4%)

116.5 722.9 42320.0 (16.8%)

– 3879.1 39760.3 (15.8%)

3999.7 (1.6%) 5414.0 (2.1%) 252160.4

PVTGC ¼

J X T X t X S X DFðtÞ  FPj;t  r j;v  Pj;t;v ;s  hs j¼1 t¼1 v ¼0 s¼1

þ

J X T X t X S X

DFðtÞ  VAROM j;t;v  P j;t;v ;s  hs

means that fuel/output ratio decreases as vintage increases (fuelsaving technological progress). By analogy, suppose that FPj,t also takes place at a given exponential growth rate with a residual [22–24]:

j¼1 t¼1 v ¼0 s¼1 J X J X T T X X þ DFðtÞ  FIXOMj;t  CAPj;t þ DFðtÞ  C j;t  X j;t j¼1 t¼1

j¼1 t¼1

where parameter j is generation technology (1. . .J), or type of fuel used to generate electricity since technology in the proposed model is categorized by type of fuel; t is planning period (1. . .T); v is vintage (0. . .V) with 0 for existing power plants and 1. . .V for newly installed power plants during the planning period; s is the block (1. . .S) formed by the time axis on the load duration curve.

FbP jtþ

t X

FPj;t ¼ FPj;0  e

j eFP; t

t¼1

c j is conwhere FPj,0 is the initial price of fuel j; eFP;j is the residual; FP stant growth rate. The positive value of the growth rate means that fuel price will increase over time. In addition, future capital cost estimates are established by the one-way learning curve. Investment cost reductions per unit installed capacity is dependent on cumulative installed capacity. The residual is also included in the learning curve and expressed as follows:

DF(t): the discount factor, DF(t) = (1 + R)t. FPj,t: unit price of fuel j at period t. rj,v: amount of fuel j required for a given level of power output from power plants installed in vintage v; Pj,t,v,s: power output of technology j from power plants installed in vintage v during period t in block s; hs: duration time of block s; VAROMj,t,v: variable O&M cost of technology j of power plants installed in vintage v during period t. FIXOMj,t: fixed O&M cost of technology j of power plants installed in vintage v during period t; CAPj,t: the cumulative installed capacity of technology j at period t; Cj,t: the capital cost per unit of capacity of technology j at period t; Xj,t: the new installed capacity of technology j at period t; Assuming that fuel consumption rate per unit of power output is influenced by vintage only, the higher the generation efficiency of new power plants, the less fuel required. Thus the technological change of fuel consumption rate (rj,v) takes place at a given exponential growth rate and with a residual [22–24]:

a

C j;t ¼ C j;0  ðnt Þ  e

r j;v ¼ r j;v1  e

t X ¼ C j;0  X j;v

!a

C;j

eet

v ¼1

where Cj,0 is the initial capital cost for technology j; eC;j is the residual; nt is the cumulative installed capacity; and a is the learning parameter. All error terms (er;j ; eFP;j ; eC;j ) are assumed to have zero expectation, constant variance, and serially uncorrelated. By assumptions, 2 x;j we have (Eðex;j Þ ¼ 0Þ; Eððex;j Þ Þ ¼ r2x ; Eðex;j w  ew1 Þ ¼ 0 for x = r, FP, C; w = v, t. 4.1.2. The risk of total generation costs (the variance of total generation costs) Because VarðPVTGCÞ ¼ E½PVTGC  EðPVTGCÞ2 By adding the residual into the proposed model, PVTGC can be rewritten. From the above assumption (Eðer;j Þ ¼ 0; EðeFP;j Þ ¼ 0; EðeC;j Þ ¼ 0), we can calculate E(PVTGC), using PVTGC and E(PVTGC). Such that complex calculation4 obtains J X J X T X T X

1;j ;j

1;j2 1 2 1 minðt1; t2Þ  B1;j t 1  Bt 2  rt 1 ;t2

j1 ¼1t 1 ¼1j2 ¼1t2 ¼1

ðk ¼ 1Þ

j ^r jþer; v

eC;j t

þ

J X J X T X T X

2;j ;j

2;j2 1 2 1 minðt1; t2Þ  B2;j t 1  Bt 2  rt 1 ;t 2

j1 ¼1t 1 ¼0j2 ¼1t 2 ¼0 r; j

r j;v1 ¼ r j;v2  e^rjþev1

ðk ¼ 2Þ

þ

J X J X T X T X

3;j ;j

3;j2 1 2 1 B3;j t 1  Bt2  rt 1 ;t2

ðk ¼ 3Þ

j1 ¼1t 1 ¼1j2 ¼1t 2 ¼1

 B1;j t ¼

^rjþer; j 1

r j;1 ¼ r j;0  e

^rjvþ

r j;v ¼ r j;0  e

b DFðtÞ  FPj;0  r j;0  e FP jtþ^r jv  P j;t;v;s  hs ;

v¼0 s¼1

Such that continuous iteration obtains v P

t X S X

er;v j

B2;j t ¼

T X S X v¼t

b DFðvÞ  FPj;0  r j;0  e FPjvþ^rjt  Pj;v;t;s  hs ;

s¼1

v¼1

where rj,0 is the initial fuel/output ratio of fuel j; er;j is the residual; r j is constant growth rate. The negative value of the growth rate

4 The detailed derivations of the risk of total generation costs could be found in Huang and Wu [22].

281

J.-H. Wu, Y.-H. Huang / Applied Energy 119 (2014) 278–287

B3;j t ¼ DFðtÞ  C j;0 

t X

!a X j;v

previous period plus the capacity of newly installed plants minus the capacity of retired plants. Constraint 4: Peaking reserve constraint

 Xj;t :

v ¼1

where min(t1, t2) represents the minimum of t1 and t2. If j1 = j2 then rjt11;j;t22 is the variance; otherwise rjt11;j;t22 is the covariance. K refers to the different types of risks considered in the model: FP is the risk (fluctuation) of fuel price growth (k = 1); r is the risk of technological progress of fuel consumption rate (k = 2); and C is the risk of capital cost reduction (k = 3). After combining the present value of total generation costs and the risk of total generation costs, we can obtain ‘‘the present value of total generation costs after risk adjustment’’. That is,

J X CAPj;t P Dt;1  ð1 þ mÞ ðs ¼ 1; t ¼ 1 . . . . . . TÞ j¼1

where Dt,1 is the load demand during peak period (s = 1), and m denotes the reserve margin. This constraint ensures that total installed capacity satisfies the load demand of the peaking time-slice (s = 1) by a certain percentage (reserve margin). Constraint 5: Capacity constraint

Pj;t;v ;s 6 aj  X j;v

RAðPVTGCÞ ¼ PVTGC þ k  VarðPVTGCÞ k is the risk-averse parameter. It also represents the relative contribution of the variance of total generation costs in the objective function. If k is zero then the risk of total generation costs is excluded from the technology portfolio selection. The higher the value is, the more risk-averse the investor is. The objective function, based on the above, is Min RA(PVTGC).

Constraint 1: Balance equation between power supply and demand

ðs ¼ 1 . . . . . . S; t ¼ 1 . . . . . . TÞ

v ¼0

In the formula above, Dt,s represents the power demand during period s of year t, and Losst denotes the sum of transmission and distribution loss rates. This constraint ensures that total power output must satisfy the load demand of each period after the deduction of line losses. Constraint 2: Operation constraint If electricity generating technology j is coal-fired or nuclear, and s is at peak demand (s = 1), then

Pcoal;t;v ;1 ¼ 0; Pnuclear;t;v ;1 ¼ 0 Power output is limited by unit characteristics. Base load units (e.g. coal-fired, nuclear) can operate stably for an extended period. They normally have a higher fixed cost and lower variable cost. Peak load units (e.g. LNG-fired) can be switched on and off quickly to provide power. They have lower fixed cost and a higher variable cost. Middle load units are between the base load and peak load properties. This constraint restricts a unit’s power supply capabilities because of operational properties. Coal-fired and nuclear units have a limited load adjustment speed and are unable to undertake the peak load. To conform to these operational constraints, their peak load power generation is set to zero to reflect the actual situation [25,26]. Constraint 3: Capacity transfer

CAPj;t ¼ CAPj;t1 þ Xj;t  Retirej;t

aj is the availability factor of technology j. This constraint ensures that the actual power output from different technologies does not exceed their available capacity during each period. Constraint 6: Power generation constraint of renewable energy technologies S X Pj;t;v ;s  hs 6 8760  bj  X j;v

ðv ¼ 0 . . . t; t ¼ 1 . . . T; j

2 renewable energy technologiesÞ

This model is subject to 11 constraints, which are explicated below with the aid of mathematical models.

j¼1

¼ 1 . . . . . . JÞ

s¼1

4.2. Constraints

J X t X Pj;t;v ;s  ð1  Losst Þ P Dt;s

ðs ¼ 1 . . . . . . S; v ¼ 0 . . . . . . t; t ¼ 1 . . . . . . T; j

ðj ¼ 1 . . . . . . J; t ¼ 1 . . . . . . TÞ

where Retirej,t denotes the capacity of technology j that retire in year t. This formula calculates cumulative installed capacity of different technology types during each period. The cumulative installed capacity equals the cumulative installed capacity of the

bj represents the capacity factor of renewable energy technology. Due to seasonal and climatic factors that affect the reliability of renewable energy, the capacity factors of renewable energy technologies are relatively lower than those of fossil fuels. The constraint must be placed on capacity factors to limit their maximum power output. Constraint 7: Capacity constraint of LNG reception terminals t X S X r gas;v  Pgas;t;v ;s  hs 6 gasimportt ðt ¼ 1 . . . . . . T; j 2 gasÞ

v ¼0 s¼1

gasimportt denotes the available supply of liquefied natural gas (LNG) for power generation each year. This constraint ensures that the amount of LNG used does not exceed the available supply, the imported amount of which is limited by the receiving capacity of reception terminals. Constraint 8: Carbon dioxide emissions relationship J X t X S X ej  r j;v  Pj;t;v ;s  hs ¼ CO2t ðt ¼ 1 . . . . . . TÞ j¼1

v ¼0 s¼1

ej is the carbon emissions coefficient of fuel used in technology j. This formula calculates the amount of carbon dioxide emissions during each planning period and can be adjusted to limit carbon dioxide emissions in some selected scenarios. Constraint 9: Potential limitation of renewable energy technologies

CAPj;T 6 MaxX j

ðj 2 renewable energy technologiesÞ

MaxXj denotes the development potential of renewable energy technology j. This formula defines the upper limit of development in renewable energy technologies. Although renewable energies are produced locally, the development potential is also subject to geographical conditions. This constraint ensures that the development capacity of renewable energies does not exceed the potential. Constraint 10: Non-negative constraint

Pj;t;v ;s P 0 ðs ¼ 1 . . . . . . S; v ¼ 0 . . . . . . t; t ¼ 1 . . . . . . T; j ¼ 1 . . . . . . JÞ The specified variables must be zero or positive. Constraint 11: Intermittency constraint of renewable energies

P wind;t;v ;peak 6 CAP wind;t  capacity creditwind ðv ¼ 0.... ..t;t ¼ 1..... .TÞ

282

J.-H. Wu, Y.-H. Huang / Applied Energy 119 (2014) 278–287

CAPwind,t signifies the cumulative installed capacity of wind power during each period, and capacity creditwind is the capacity credit of wind power. This concludes our model formulation. The overall objective function includes two sub-objectives: minimizing the present value of total generation costs and minimizing the risk of total generation costs. Both are linked by the risk-adverse parameter. In addition to the objective function, we also implemented 10 constraints, to which an 11th constraint was added in consideration of the intermittency of renewable energies. There are no countryspecific constraints included in the model formulation. It means that the model can be universally applied to other countries and regions, provided that the necessary data for model calibration are available. In the following, we describe the data sources and adjustments of the model for the case of Taiwan. 5. Data sources and adjustments As an electricity portfolio planning model, the model constructed in this paper focused on the technology portfolio selection of the supply side; the load demand is assumed to be exogenously given. Furthermore, we considered the characteristics of six types of renewable energy technologies: conventional hydro, wind, solar PV, municipal solid waste (MSW), other biomass, and geothermal energy. In addition to renewable energies, the model also considered three types of fossil fuel technologies (coal-fired, oil-fired, and LNG-fired) and nuclear technology. 5.1. Fuel price, fuel/output ratio and capital cost The technical and economic parameters for ten types of power plants are summarized in Table 2. Fuel purchase price is not made public as it is power company classified information. Hence, fuel price is based on data released by the Bureau of Energy (average purchase prices for imported energy, 1990–2010) [27]. The historical data on fuel price are deflated to year 2006 currency value over the designated period and used to estimate growth rate of the proposed model function (Table 2). Fuel price estimation in future planned years is made on the basis of fuel price. Fuel/output ratio (fuel consumption rate) is developed from data released by the Bureau of Energy [27]. The amount of fuel required per unit of power output is calculated by dividing power plant fuel consumption by total electricity generation. Fuel/output

ratio of MSW is calculated from data released by the Environmental Protection Administration [28]. The fuel/output ratio rate for MSW also applies to other biomass energy due to a lack of data regarding fuel consumption and electricity generation. The historical data on fuel/output ratio have been used to estimate growth rate of the proposed model function (Table 2). Fuel/output ratio estimation in future planned years is made on the basis of fuel consumption growth rate (a negative value implies fuel-saving technological progress). The learning curve forms the basis of estimated future capital cost in the model. Cumulative installed capacity determines reduction rate in capital cost per unit of installed capital. Only renewable energy technologies are assumed to exhibit learning effects. However, due to cost and cumulative installed capacity data merely available for wind and solar PV technologies in Taiwan, estimations are made on the two technologies (Table 2). Estimated future target year capital cost is made on the basis of learning rate. 5.2. The variance and covariance The risk of changes in fuel price, fuel/output ratio and capital cost is calculated based on data from the Bureau of Energy and Environmental Protection Administration from 1990 to 2010 (Table 2). In terms of fuel price fluctuation risk (i.e. variance for fuel price growth), oil is the most risky resource followed by LNG, coal, and nuclear. The fuel price fluctuation risk for conventional hydro, wind, solar PV and geothermal technology are zero as no fuel is required. The fuel price fluctuation risk for MSW and other biomass are zero as only waste is involved. 5.3. Load demand data and line loss rate Historical and future peak load demands are shown in Table 3. Peak load demand is expected to grow from 37334 MW in 2011 to 53582 MW in 2025 with an average growth rate of 2.61% [29]. The Taipower Planning Department estimates that line loss rate is projected to fall from 4.72% in 2011 to 4.52% in 2025 due to line loss improvement plans. 5.4. Upper limit for LNG import Yong-an and Taichung are currently two LNG reception terminals in Taiwan with total receiving capacity of 12,000 thousand

Table 2 Technical and economic parameters for ten types of power plants [27,28].

a b c d e

Parameters technologies

Initiala fuel priceb (NTD)c

Initial fuel/ output ratiod

Initial capital coste

Growth rate of fuel price

Growth rate of fuel/output ratio

Learning rate

Variance for fuel price growth (r2FP )

Variance for fuel consumption rate (r2r )

Variance for capital cost reduction (r2C )

Coal-fired Oil-fired LNG-fired Nuclear Conventional hydro Wind Solar PV MSW Other biomass Geothermal energy

3850 21917 15,530 79.94 0

0.3889 0.2477 0.1942 2.4021 0

64.3 32.7 28.3 124.3 90

0.0252 0.0453 0.0305 0.0162 0

-0.00047 -0.00124 -0.01748 -0.01282 0

0 0 0 0 0

0.034942 0.053488 0.040359 0.024025 0

0.002750 0.000315 0.000866 0.001719 0

0 0 0 0 0

0 0 0 0

0 0 2.0606 2.0606

78.2 152.6 150.4 98.1

0 0 0 0

0 0 -0.03913 -0.03913

0.025 0.120 0 0

0 0 0 0

0 0 0.001957 0.001957

0.003988 0.000553 0 0

0

0

233.1

0

0

0

0

0

0

Initial represents year 2011. The unit of initial fuel price: coal (NTD/tons), fuel oil (NTD/kilolitres), LNG (NTD/thousand cubic meters), uranium (NTD/grams). NTD stands for New Taiwan Dollar; 1 USD is approximately equivalent to 30 NTD. The unit of initial fuel/output ratio: coal (tons/MWh), fuel oil (kilolitres/MWh), LNG (thousand cubic meters/MWh), uranium (grams/MWh). The unit of initial capital cost: millions NTD/MW.

J.-H. Wu, Y.-H. Huang / Applied Energy 119 (2014) 278–287

283

5.8. Other parameters

Table 3 Peak load demand and line loss rate [29]. Year

Peak load demand (MW)

Growth rate (%)

Line loss rate (%)

2011 2012 2013 2014 2015 2016 2017 2018 2019 2020 2021 2022 2023 2024 2025

37,334 38,040 39,406 40,862 42,399 43,823 45,192 46,502 47,756 48,954 49,969 50,932 51,845 52,710 53,582

– 1.89 3.59 3.69 3.76 3.36 3.12 2.90 2.70 2.51 2.07 1.93 1.79 1.67 1.65

4.72 4.70 4.69 4.67 4.66 4.64 4.63 4.61 4.60 4.58 4.57 4.55 4.54 4.52 4.52

tons per annum. Given the implementation of scheduled expansion projects, their receiving capacity should reach 15,000 thousand tons per annum in 2015, 18,000 thousand tons in 2020 and 20,000 thousand tons in 2025.

5.5. Development potential of renewable energies The capacity of renewable energy is subject to geographical conditions despite being an indigenous energy resource. Renewable energy development potential estimation in Taiwan is made based on data from 3rd National Energy Conference and the New Energy and Clean Energy Research and Development Planning Report [30,31] shown in Table 4.

5.6. Calculation of CO2 emissions There are many types of greenhouse gases; however, in this paper we primarily considered carbon dioxide emissions. Our estimation was based on carbon emissions coefficients and carbon oxidation rates released by the Intergovernmental Panel on Climate Change (IPCC), whereby we calculated the carbon dioxide emissions coefficients of each type of fuel. The coefficients were then multiplied by the amount of fuel consumed to estimate carbon dioxide emissions.

5.7. Capacity credit of renewable energy Renewable energies with intermittent power output are principally wind power and solar power. However, the total installed capacity of solar PV in Taiwan was a mere 41.1 MW (from a total of 2106 systems) at the end of 2011. With the exception of a few larger systems, the solar PV systems had capacities of 10 KW or less and were widely distributed among schools, common organizations, and private residences. Due to the difficulty in obtaining hourly operation details for the solar PV systems, we considered the capacity credit of wind power in this respect. In this paper, we used the load duration curve (presenting load patterns) of 2010 and the power output of wind power plants in 2010 (presenting the variance of wind power output) to estimate the capacity credit of wind power. However, due to the limited availability of hourly operation data, we calculated the capacity credit of wind power based on the year-round operation data collected from three representative wind sites (i.e. Mailiao, Penghu Zhongtun and Hsinchu Andante).

Reserve margin rate is set at 16% by the government. The discount rate remains constant at 5%,5 and the modeling period is 14 years, from 2012 to 2025. 6. Results 6.1. Scenario design The scenarios in our simulations are designed as follows: Case 0 (C0) is the baseline scenario used to examine annual technology portfolios under the least-cost principle (i.e. the riskaversion parameter setting is 0) and objective function without considering the influence of risk. Cases 1 through 4 (C1–C4) primarily consider the characteristics of renewable energy that hedge against fluctuating fossil fuel price risk. Using various risk-adverse parameters (set as 0.001, 0.0025, 0.005, and 0.0075), we observed the impact of risk levels on technology portfolios. Case 5 (C5) prioritizes the reduction of carbon dioxide emissions. In the constraint condition, we set an upper limit for the amount of carbon dioxide emissions to observe the impact of this objective on electricity generating portfolios. According to the Master Plan of Energy Conservation and Carbon Mitigation proposed in 2010, Taiwan’s objective is to reduce the amount of carbon dioxide emissions to that of 2000 in 2025. It is this objective at which we set upper limit of carbon dioxide emissions for the simulation. Due to its influence on power supply stability, the intermittent nature of some renewable energy must be taken into consideration to help mitigate its impact. The model that we constructed in this paper is therefore a long-term planning with a focus on capacity credit in terms of renewable energy intermittency. The estimation results indicate that the capacity credit of wind power is 0.18, which is only 60% of the annual capacity factor (i.e. 0.3). This shows that wind power in Taiwan has a weaker power output during annual peak demand (i.e. summer) but can generate more electricity during off-peak periods (i.e. winter). Therefore, wind power is not a reliable option during peak demand, and during this period, the actual output of wind power should be limited to below the maximum credit of peak load to minimize the influence of wind power intermittency on the entire power system. Once the proportion of renewable energies in the power system increases, their capacity credit will gradually decline. For this reason, we included another case considering the intermittency of renewable energies with a lower capacity credit of 0.12 for wind power (Case 7), in addition to the case with a capacity credit of 0.18 (Case 6). In summary, Cases 6 and 7 (C6–C7) primarily use estimated credit during peak load to determine the influence of intermittency on technology portfolios and total generation costs. The case designs above investigated three major scenarios involving the influence of generation cost-related risks, the reduction of carbon dioxide emissions, and the intermittency of renewable energy on technology portfolios and total generation costs. The parameter settings for each case are summarized in Table 5. 6.2. Simulation results In the latest goals for renewable energy development set by the government of Taiwan, the total installed capacity of renewable energies is aimed for 9952 MW in 2025, accounting for 14.8% of 5 The discount rate (5%) was determined by employing actuarial methods to calculate the most recent 10-year treasury rate (1.3%), which represents the risk-free interest rate and risk premium (3.7%).

284

J.-H. Wu, Y.-H. Huang / Applied Energy 119 (2014) 278–287

Table 4 The development potential of renewable energies in Taiwan [30,31]. Technologies

The development potential of renewable energies

Conventional hydro Wind

Total development potential is estimated to be 4.12 GW

MSW Other biomass Geothermal energy

the total capacity [32]. Since the goal for wind power is 3000 MW which claims the largest proportion in the total renewable energy capacity. We therefore explain the results of the simulations primarily from the aspect of wind power. Fig. 1 illustrates the impact of different risk-averse parameters (C0–C4) on technology portfolios using wind power as an example. It indicates that the greater the degree of risk aversion, the greater the proportion of installed capacity in wind power. Furthermore, installed capacity is increased earlier to replace fossil fuel technologies with risk-aversion level increase. For the baseline scenario (C0, without considering the risk aversion), newly installed capacity was not added until the 4th period (the year 2015). In Case 4 (C4), the scenario with the highest risk aversion, installed capacity was increased in the 1st period (the year 2012). The simulation results mainly illustrate that taking the risk of total generation cost into account creates a preference for more risk-averse renewable energy in the technology portfolios and also helps reduce portfolios exposure to fossil fuel price fluctuations. However, due to the limitations of geographical conditions for wind power, we set 3000 MW as the upper limit of development potential for wind power based on the estimates provided by the Bureau of Energy of the Ministry of Economic Affairs [30,31]. Once this upper limit was achieved, installed capacity could no longer be increased. For this reason, the proportions of wind power in the simulated scenarios were the same in the latter portions of the simulation (in other words, they all reached the maximum development potential). Fig. 2 indicates that when a constraint is imposed on carbon dioxide emissions, the proportion of wind power in the total installed capacity is higher than in a scenario that does not consider the reduction target. Moreover, the installed capacity is increased in an earlier period of the planning to replace fossil fuel technologies with reduction target becoming tighter. In the scenario in which the reduction of carbon dioxide emissions is not considered (C1), newly installed capacity for wind power would not be implemented until the 3rd period (the year 2014) of the simulation. With a constraint implemented (C5), installed capacity

5.0% 4.5% 4.0% 3.5%

Share (%)

Solar PV

Total development potential for wind power is estimated to be 3 GW. However, after considering the issues such as the high cost of land acquisition only 1.2 GW from onshore sources is expected to be utilized to generate electricity. Together with 1.8 GW from offshore sources, total development potential for wind is 3 GW Total development potential is estimated to be 12 GWp including those installed at residential installations, industrial and commercial installations, public facilities, and other places Total development potential is estimated to be 0.9 GW based on the amount of urban waste Total development potential is estimated to be 0.8 GW based on the amount of industrial waste and agricultural waste Total development potential is estimated to be 0.65 GW based on data from the 25 non-volcanic geothermal areas

3.0%

C0 C1 C2 C3 C4

2.5% 2.0% 1.5% 1.0% 0.5% 0.0%

0

1

2

3

4

5

6

7

8

9

10 11 12 13 14

Period Fig. 1. Proportions of wind power in total installed capacity resulting from varying degrees of risk.

was increased in the 1st period (the year 2012). Nevertheless, due to the upper limit of development potential setting for wind power, the installed capacity could no longer be increased after this limit. Therefore, the proportions of wind power in the two scenarios (C1 and C5) were the same in the last portion of the simulation. In summary, from above the two major scenarios (considering the risk of total generation costs (C1–C4) and that which considered the reduction of carbon dioxide emissions (C5)), either higher risk aversion or tighter constraints on carbon reduction lead to an increase in the proportion of installed capacity of renewable energies as fossil fuel technologies substitutes. With these objectives, this increase is also implemented earlier in the planning period. Consequently, we can see that the benefits reaped by the use of renewable energies are twofold: a hedge against the risks of volatile fossil fuel prices and reduction of carbon dioxide emissions. These energies are thus a viable alternative for fossil fuels.

Table 5 Parameters adopted for the simulations. Scenarios considering the risk of total generation costs Scenarios Risk-aversion parameters

Case 0 (C0) 0

Case1 (C1) 0.001

Scenarios considering the reduction of carbon dioxide emissions Scenarios Upper limit for CO2 emissions in 2025 (CO22025) Risk-aversion parameters Scenarios considering the intermittency of renewable energy Scenarios Capacity credit of wind power Risk-aversion parameters

Case2 (C2) 0.0025

Case3 (C3) 0.005

Case5 (C5) 130 million tons 0.001 Case6 (C6) 0.18 0.001

Case7 (C7) 0.12 0.001

Case4 (C4) 0.0075

285

J.-H. Wu, Y.-H. Huang / Applied Energy 119 (2014) 278–287

40.0%

5.0% 4.5%

35.0%

4.0%

3.0% C1

2.5% 2.0%

C5

Share (%)

Share (%)

3.5% 30.0% C1

25.0%

C6 C7

1.5% 20.0%

1.0% 0.5%

15.0%

0.0% 0

1

2

3

4

5

6

7

8

9

10 11 12 13 14

0

1

2

3

6

7

8

9

10 11 12 13 14

Fig. 3. Proportions of LNG-fired power plants in total installed capacity resulting from different capacity credits for renewable energies.

5365 5345 5325

billion NTD

The results also show that the replacement effect (i.e. renewable energy technologies replacing fossil fuel technologies) is limited as geographical conditions dictate renewable energy development potential. However, as future power generation technologies continue to advance, the energy density of renewable energies is expected to be effectively increased (for example, increasing the capacity of individual wind turbines) to reduce the amount of land needed and extend the development limit for further replacement of fossil fuels. In the scenarios considering the intermittency of renewable energies, the simulation results revealed that the main influence on the proportion of wind power was LNG-fired power plants. Fig. 3 depicts the influence of different capacity credits for wind power (C6 and C7) on the proportion of LNG-fired power plants. The results show that when the intermittency of renewable energies is taken into consideration (C6 and C7), the proportion of installed capacity in LNG-fired technology becomes higher than when intermittency is not considered (C1). Furthermore, when wind power displays a lower credit during peak load, the proportion of LNG-fired plants is higher in the total installed capacity. The reason for this is that without consideration of intermittency, the electricity generated by wind power is primarily used to satisfy peak demand and a small portion of middle demand. When the intermittency of wind power is included, the actual power output at peak demand cannot exceed the maximum credit of the peak load, which restricts the capability for supporting peak demand of wind power (i.e. generating less electricity during peak demand). The reserve margins of other technologies are thus required to make up for the shortage. Due to their quick responsiveness to sudden load demand changes, LNG-fired plants can easily handle the increase in demand during peak periods and are the primary backup generators for wind power. During the first half of the simulation period (period 1–6), the existing installed capacity for LNG-fired plants was sufficient to meet peak demand and the backup required for wind power. After 2018, LNG-fired plants were added to meet the increasing demand for electricity; the installed capacity began to increase, and the total installed capacity of LNG-fired power plants reached 22,805 MW, occupying 32.94% of the entire electricity sector in 2025. Taking the intermittency into consideration (C6–C7), LNG-fired plants were added to make up for the capability for supporting peak demand of wind power. Nevertheless, due to the upper limit of development potential setting for wind power, the additional increase is only 1.45% in the proportion of LNG-fired plants in 2025. Fig. 4 exhibits the total generation costs in Case 1 without capacity credits consideration and in the scenarios in which the

5

Period

Period Fig. 2. Proportions of wind power in total installed capacity resulting from a constraint on carbon dioxide emissions.

4

5338.4

5340.3

C6

C7

5311.5

5305 5285 5265 5245 5225 5205 C1

Scenarios Fig. 4. Total generation costs resulting from different capacity credits for renewable energies.

capacity credits of wind power are 0.18 (C6) and 0.12 (C7), respectively. In Case 1, the intermittency of wind power is not considered; most part of wind power generation units can be used to satisfy peak demand without requiring the additional costs of backup generators. In the other two scenarios, when intermittency is included, the capability of wind power to support peak load weakens, thereby requiring LNG-fired plants to make up the shortage. This would increase the capital as well as the operation and maintenance cost for the backup generators (LNG-fired technology). However, wind power can only take up 4.6% of the total installed capacity, and only an additional 1.45% of installed capacity was required for LNG-fired plants until 2025. As a result, the total generation cost in Case 6 is NTD 26.9 billion more than in Case 1. Furthermore, as the capacity credit of wind power declines (Case 7), the total cost increases. The simulation result shows that the total cost of Case 7 is NTD 1.9 billion more than that on Case 6. In summary, the intermittency of renewable energies will restrict the capability of the wind power to support peak loads, requiring other technologies (such as LNG-fired technology) to serve as the reserve margin and make up for the shortage, which in turn will increase total generation costs. However, because the current upper limit for wind power is set at 3000 MW, the influence of intermittency on the reserve margins of the power system and the total generation costs is limited.

7. Conclusion Traditional electricity planning models apply the least-cost method to select from a range of electricity generating

286

J.-H. Wu, Y.-H. Huang / Applied Energy 119 (2014) 278–287

technologies. This approach disregards many characteristics of renewable energy technologies. As an indigenous energy, the correlation between the price volatility of renewable energy and that of fossil fuels is weak. Therefore, increasing the proportion of renewable energy in the power generation structure can spread the volatility risk of imported fuels. In addition, renewable energy technologies, due to their greater potential for cost reductions, are expected to successfully compete with conventional technologies at some time in the future. On the other hand, renewable energies are affected by the time of the day, the season and the weather, and this intermittency will surely affect the continuity and stability of the power supply. Thus, the problem of integrating different features of renewable energies into traditional electricity planning model is crucial. This paper attempts to apply portfolio theory, learning curve theory, and the capacity credit to consider characteristics of renewable energy, such as a hedge against fossil fuel price volatility, significant technological progress, and intermittent generation. The relevant constraints of traditional electricity planning models are also incorporated in model construction. Finally, the proposed model is implemented for the case study of Taiwan’s electricity sector to illustrate the impact of various influential factors on technology portfolios and total generation costs. The simulation results of the scenarios considering the risks of total generation costs show that the greater the degree of risk aversion, the greater the proportion of installed capacity in wind power. Furthermore, taking into account these risks also means installed capacity would be increased at a more rapid rate to replace fossil fuel technologies. Similarly, the results of the scenario considering the reduction of carbon dioxide emissions show that the use of renewable energy contributes greatly to reducing carbon dioxide emissions. The tighter the reduction goal is, the greater the extent of replacement of fossil fuel technologies with renewable energies. Therefore, the results demonstrate that using renewable energies has the advantage of hedging against the volatile fossil fuel price risk as well as reducing carbon dioxide emissions. However, the results also show that the replacement effect is limited as geographical conditions dictate renewable energy development potential. As future power generation technologies continue to advance, the energy density of renewable energies is expected to be effectively increased to reduce the amount of land needed and extend the development limit for further replacement of fossil fuels. Considering the intermittency of renewable energies restricted the capability of wind power to support peak loads, requiring LNG-fired power plants to serve as the reserve margin and make up for the shortage. This increased the total generation costs. With the upper limit of 3000 MW for wind power and the highest resulting proportion of 4.6% in the total installed capacity, taking the intermittency of wind power into account led to a slight additional increase of 1.45% in the proportion of LNG-fired plants in 2025 as well as a generation cost increase of NTD 26.9 billion. The significance of our research is that it attempts to introduce different complementary approaches to traditional electricity planning model. Embodied technical progress is also considered in the model which means the technical progress displayed by newer and more efficient capital performance. Under these circumstances, the capital structure is no longer homogeneous and comprises different vintages. There are no country-specific constraints included in the model formulation. It means that the model can be universally applied to other countries and regions, provided that the necessary data for model calibration are available. Therefore, the benefit of renewable energy technologies in reducing the generation cost-related risks and carbon dioxide in the electricity sector can be better identified based on a theorysupported framework as demonstrated herein.

From above simulation results, some policy suggestions are provided. Firstly, because of the low energy density and intermittent generation of renewable energy, it is currently unfeasible to replace fossil fuels on a very large scale; therefore budgets associated with the development of energy technologies should be increased to promote further research on renewable energy technologies with high unit capacity, high efficiency, and stable supply. Secondly, alliances between R&D and industry should be forged and incentive grants established to encourage up-, mid-, and downstream companies and research institutes in the cooperation and development of the local renewable energy industry. The model developed in this paper only considered some characteristics of renewable energy, including a hedge against fossil fuel price volatility, significant technological progress, and intermittent generation, whereas the value of hedging against uncertain carbon prices was not addressed. The important issue has been highlighted in many studies [33–42]. Although Taiwan does not implement emissions trading system at present, the scheme could still play an important role to reduce CO2 emissions in the future. Future research could cover the carbon price issue and give consideration to the value of hedging against uncertain carbon prices in this model. Acknowledgment The authors would like to thank the National Science Council of the Republic of China, Taiwan, for financially supporting this research under Contract No. NSC-100-2410-H-006-072-. References [1] Pode R. Addressing India’s energy security and options for decreasing energy dependency. Renew Sustain Energy Rev 2010;14:3014–22. [2] Huang YH, Wu JH. Energy policy in Taiwan: historical developments, current status and potential improvements. Energies 2009;2:623–45. [3] Bhattacharya A, Kojima S. Power sector investment risk and renewable energy: a Japanese case study using portfolio risk optimization method. Energy Policy 2012;40:69–80. [4] International Energy Agency (IEA) Experience curves for energy technology policy. Paris: IEA; 2000. [5] McDonald A, Schrattenholzer L. Learning rates for energy technologies. Energy Policy 2001;29:255–61. [6] Berglund C, Söderholm P. Modeling technical change in energy system analysis: analyzing the introduction of learning-by-doing in bottom-up energy models. Energy Policy 2006;34:1344–56. [7] Söderholm P, Sundqvist T. Empirical challenges in the use of learning curves for assessing the economic prospects of renewable energy technologies. Renew Energy 2007;32:2559–78. [8] Schoots K, Ferioli F, Kramer GJ, van der Zwaan BCC. Learning curves for hydrogen production technology: an assessment of observed cost reductions. Int. J. Hydrogen Energy 2008;33:2630–45. [9] Pettersson F, Söderholm P. The diffusion of renewable electricity in the presence of climate policy and technology learning: the case of Sweden. Renew Sustain Energy Rev 2009;13:2031–40. [10] Plaza M, Ngwenyama OK, Rohlf K. A comparative analysis of learning curves: implications for new technology implementation management. Eur J Oper Res 2010;200:518–28. [11] Yu CF, van Sark WGJHM, Alsema EA. Unraveling the photovoltaic technology learning curve by incorporation of input price changes and scale effects. Renew Sustain Energy Rev 2011;15:324–37. [12] Li S, Zhang X, Gao L, Jin H. Learning rates and future cost curves for fossil fuel energy systems with CO2 capture: methodology and case studies. Appl Energy 2012;93:348–56. [13] Zhu L, Fan Y. Optimization of China’s generating portfolio and policy implications based on portfolio theory. Energy Policy 2010;35:1391–402. [14] Gökgöz F, Atmaca ME. Financial optimization in the Turkish electricity market: Markowitz’s mean-variance approach. Renew Sustain Energy Rev 2012;16:357–68. [15] Allan G, Eromenko I, McGregor P, Swales K. The regional electricity generation mix in Scotland: a portfolio selection approach incorporating marine technologies. Energy Policy 2011;39:6–22. [16] Arnesano M, Carlucci AP, Laforgia D. Extension of portfolio theory application to energy planning problem- the Italian case. Energy 2012;39:112–24. [17] Muñoz JI, Sánchez de la Nieta AA, Contreras J, Bernal-Agustín JL. Optimal investment portfolio in renewable energy: the Spanish case. Energy Policy 2009;37:5273–84.

J.-H. Wu, Y.-H. Huang / Applied Energy 119 (2014) 278–287 [18] Awerbuch S, Berger M. Applying portfolio theory to EU electricity planning and policy-making. Paris: IEA; 2003. [19] Awerbuch S, Yang S. Efficient electricity generating portfolios for Europe: maximizing energy security and climate change mitigation. EIB Papers 2007;12:8–37. [20] Chang SL. The analysis of the optimal electricity energy mix decision in Taiwan: application of MULTEEE model. In: Proceedings. The 30th International Conference of the International Association for Energy Economics, Wellington, New Zealand; 2007. [21] Ko FK, Huang CB, Tseng PY, Lin CH, Zheng BY, Chiu HM. Long-term CO2 emissions reduction target and scenarios of power sector in Taiwan. Energy Policy 2010;38:288–300. [22] Huang YH, Wu JH. A portfolio risk analysis on electricity supply planning. Energy Policy 2008;36(2):627–41. [23] van Zon A, Fuss S. Investing in energy conversion technologies – an optimum vintage portfolio selection approach. Maastricht Economic Research Institute on Innovation and Technology (MERIT); 2005, Working Paper 2005-013. [24] van Zon A, Fuss S. Irreversible investment under uncertainty in electricity generation: a clay-clay vintage portfolio approach with an application to climate change policy in the UK. Maastricht Economic Research Institute on Innovation and Technology (MERIT); 2006, Working Paper 2006-035. [25] Ko L, Chen CY, Lai JW, Wang YH. Abatement cost analysis in CO2 emission reduction costs regarding the supply-side policies for the Taiwan power sector. Energy Policy 2013;61:551–61. [26] Ko L, Chen CY, Seow VC. Electrical power planning and scheduling in Taiwan based on the simulation results of multi-objective planning model. Int J Electr Power Energy Syst 2014;55:331–40. [27] Bureau of Energy (BOE), Ministry of Economic Affairs. Energy statistics monthly. Taiwan, ROC; September 2012. [28] Environmental Protection Administration (EPA), Executive Yuan. Taiwan environmental conservation annual report for 2011. Taiwan, ROC; 2012.

287

[29] Taiwan Power Company (Taipower). Long-term load forecasting and longterm power development programming. Taiwan, ROC; 2012. [30] Bureau of Energy (BOE), Ministry of Economic Affairs. The data of 3rd National Energy Conference. Taiwan, ROC; 2009. [31] Bureau of Energy (BOE), Ministry of Economic Affairs. New energy and clean energy research and development planning report. Taiwan, ROC; 1999. [32] Bureau of Energy (BOE), Ministry of Economic Affairs. New energy policy. Taiwan, ROC; 2011. [33] Zhou Y, Wang L, Mccalley James D. Designing effective and efficient incentive policies for renewable energy in generation expansion planning. Appl Energy 2011;88(6):2201–9. [34] Tolis AI, Rentizelas AA. An impact assessment of electricity and emission allowances pricing in optimised expansion planning of power sector portfolios. Appl Energy 2011;88(11):3791–806. [35] Ghadikolaei H, Tajik E, Aghaei J, Charwand M. Integrated day-ahead and hourahead operation model of discos in retail electricity markets considering DGs and CO2 emission penalty cost. Appl Energy 2012;95:174–85. [36] Reuter WH, Szolgayová J, Fuss S, Obersteiner M. Renewable energy investment: policy and market impacts. Appl Energy 2012;97:249–54. [37] Foley A, Tyther B, Calnan P, Gallachóir BÓ. Impacts of electric vehicle charging under electricity market operations. Appl Energy 2013;101:93–102. [38] García-Martos C, Rodríguez J, Sánchez MJ. Modelling and forecasting fossil fuels CO2 and electricity and their volatilities. Appl Energy 2013;101:363–75. [39] Lam JCK, Woo CK, Kahrl F, Yu WK. What moves wind energy development in China? Show me the money! Appl Energy 2013;105:423–9. [40] Jirutitijaroen P, Kim S, Kittithreerapronchai O, Prina J. An optimization model for natural gas supply portfolios of a power generation company. Appl Energy 2013;107:1–9. [41] Mari C. Hedging electricity price volatility using nuclear power. Appl Energy 2014;113:615–21. [42] Woo CK, Sreedharan P, Hargreaves J, Kahrl F, Wang J, Horowitz I. A review of electricity product differentiation. Appl Energy 2014;114:262–72.