Applied Trade-o ﬀ s between electri ﬁ cation and climate change mitigation: An analysis of the Java-Bali power system in Indonesia

(cid:129) The Indonesian power sector pathways for meeting the Paris target are proposed. (cid:129) Meeting electri ﬁ cation goals under current energy mix triples the CO 2 emissions. (cid:129) The proposed CO 2 mitigation scenarios satisfy the Paris climate target. (cid:129) Total costs under climate mitigation e ﬀ orts equal 0.1% of GDP. (cid:129) Cost-e ﬀ ectiveness of the CO 2 mitigation scenarios is 14.9 – 41.8 US$/ton CO 2 e. The power sector in many developing countries face challenges of a fast-rising electricity demand in urban areas and an urgency of improved electricity access in rural areas. In the context of climate change, these development needs are challenged by the vital goal of CO 2 mitigation. This paper investigates plausible trade-o ﬀ s between electri ﬁ cation and CO 2 mitigation in a developing country context, taking Indonesia as a case study. Aligned with the 2015 Paris Agreement, the Government of Indonesia has announced its voluntary pledge to reduce 29% of its GHGs emissions against the business as usual scenario by 2030. 11% of this should be attained by the energy sector. We incorporate the Indonesian Paris pledge into the modelling of capacity expansion of the Java- Bali power system, which is the largest power system in Indonesia. The LEAP model is used for the analysis in this study. Firstly, we validate the LEAP model using historical data of the national electricity system. Secondly, we develop and analyse four scenarios of the Java-Bali power system expansion from the base year 2015 through to 2030. These include a reference scenario (REF) to re ﬂ ect a continuation of the present energy mix (REF), then a shift from coal to natural gas (NGS) (natural gas), followed by an expansion of renewable energy (REN) and, ﬁ nally, the least-cost option (OPT). The shift to natural gas decreases future CO 2 emissions by 38.2 million ton, helping to achieve the CO 2 mitigation target committed to. Likewise, an escalation of renewable energy development in the Java-Bali islands cuts the projected CO 2 emissions by 38.9 million ton and, thus, assures meeting the target. The least-cost scenario attains the targeted emission reduction, but at 33% and 52% lower additional costs compared to NGS and REN, respectively. The cost-e ﬀ ectiveness of CO 2 mitigation scenarios range from 14.9 to 41.8 US$/tCO 2 e.


Introduction
Electricity is a basic need in modern societies. Global electricity demand in the period of 2002-2012 increased 3.6% annually, exceeding the annual population growth for the same period [1]. Yet, 1.3 billion people worldwide still do not have access to electricity, making the provision of universal access to electricity an important The Paris Agreement requires all parties to communicate their Intended Nationally Determined Contributions (INDCs). Around 99% of the communicated INDCs cover the energy sector [4]. Accordingly, they need to incorporate their Paris target into national energy planning. Developing countries, in particular, need to align the Paris Agreement target with their vital national goals of nationwide electrification. This article addresses the question of if and how developing countries may satisfy the growing electricity demand while still meeting climate mitigation targets. We focus on Indonesia; a country with a stable growing GDP and a population of 261 million, spread across more than 13,000 islands. Aligned with the 2015 Paris Agreement, Indonesia aims to reduce its GHGs by up to 29% against business as usual, by 2030. Over and above this, an additional 12% reduction is intended with international cooperation 1 . In the meantime, 7.7 million Indonesian households still do not have access to electricity 2 . This article considers both objectives of electrification and climate change mitigation in the simulation of capacity expansion in the largest power system in Indonesia. We focus on the Java-Bali interconnected power system, which generated 74% of the national electrical energy in 2015.
This article analyses various scenarios of future power generation in the Java-Bali power system between 2016 through to 2030. We employ the Long-range Energy Alternatives Planning System (LEAP) and a unique dataset from the national electricity company (PLN). LEAP is selected over other software tools to suit our modelling needs, through a systematic screening process. Despite the fact that LEAP is actively usedin the 85 UNFCCC country report [5] and in more than 70 peerreviewed journal papers [6] publications explicitly discussing the LEAP model validation are limited. In this study, we first setup the Indonesian LEAP model and run it from the base year 2005 through to 2015. Then, we validate the model results against the historical data of the national capacity addition, electricity production and CO 2 emission over the period of 2006 through to 2015. Secondly, we develop scenarios for future power generation in the Java-Bali power system and analyse the changes in resource utilization and technology deployment that respond to the Paris pledge with eleven power generation alternatives, namely: ultra supercritical coal (USC coal), natural gas combined cycle (NGCC), natural gas open cycle (NGOC), hydro, mini hydro, hydro pumped storage, geothermal, solar photovoltaic (PV), wind power, nuclear and biomass.
The article adds a number of innovative contributions to the body of the energy modelling literature. Firstly, to the best of our knowledge, this article is the first to analyse scenarios of power system expansion, which take into account the energy sector's actual CO 2 mitigation targets associated with the Indonesian pledge to the Paris Agreement. We assess the consequences of the climate mitigation policy on the Indonesian power sector using a validated model and zoom into the level of individual technologies (a bottom-up approach), rather than employing a macro-economic approach. Secondly, we use a unique dataset from the national power company (PLN) that represents historical technical performances of every individual power plant in the Java-Bali power system of 64 plants. Such a dataset enables more accurate settings for some of the crucial LEAP input parameters, including each plant's net capacity and capacity factor 3 . Thirdly, this article is transparent on the LEAP validation procedure by using 10 years of Indonesian electricity supply and demand data. As such, the article lays out an easy-to-replicate method for assessing power sector pathways with regard to the Paris Agreement in other developing countries.
The remainder of this paper is organised as follows: Section 2 is the Literature Review; Section 3 presents the methodology and data, including validation of the LEAP model and scenarios development of the Java-Bali power system expansion; Section 4 provides the results of LEAP simulations; and Section 5 concludes the discussion.

Overview of the Indonesian power sector
Indonesia is one of the world's fast developing economies, where annual growth has averaged 5.9% from the year 2010 through to 2014 4 . Accordingly, the Indonesian power sector faces fast-growing energy demands. The average growth rate of electricity consumption in that period was 7.8% [8]. Yet, the present electrification ratio 5 in Indonesia is only 88% [9]. The country's development agenda aims to improve electrification to nearly 100% by 2020 6 . Furthermore, the Government of Indonesia is promoting an average of 5% economic growth per annum to reduce the poverty rate below 4% by 2025 7 . This implies that the demand for electricity in Indonesia will continue to grow in the next decade and is expected to more than double by 2025 [8].
By the end of 2015, the capacity of national power generation had reached 53 GW (GW) while in the same year, the peak demand was recorded at 34 GW [10]. It is pertinent to note that while the physical infrastructure spreads throughout the Indonesian archipelago, most power generation capacity (64%) is situated in the Java and Bali islands. PLN, the national power company, owns 76% of the national power generation capacity; private companies own the rest. Furthermore, the transmission and distribution grids are run solely by PLN.
The Indonesian power sector, including the Java-Bali interconnected power system, is highly dependent on fossil fuels, primarily on coal. In 2015, fossil fuels constituted 90% and 91% of the national and Java-Bali generation mixes, respectively (Fig. 1). As such, the Java-Bali power system is illustrative of the national energy mix. Furthermore, since Java and Bali are the most populated and developed islands in Indonesia, the electricity consumption in these islands continues to increase with an annual average growth of 7.5% between 2010 and 2014 [11]. In 2015, the Java-Bali power system generated 74% of the national electricity production. Consequently, Java-Bali contributes the highest share of power sector GHG emissions into the national balance, when compared to the outer islands. This article focuses entirely on the analysis of the Java-Bali power system since this is representative in terms of demand growth, energy mix and CO 2 emissions, and given our access to high quality data for this system.
Indonesia owns abundant energy resources including oil, coal, natural gas, solar, hydro, and geothermal (Tables 1 and 2). In 2015, the cumulative reserves of the three fossil fuel resources constituted 93 billion ton oil equivalent/toe. Nonetheless, the fossil fuel resources are depleting. Unless new reserves are discovered Indonesian oil are expected to be exhausted in 12 years, natural gas in 33 years, and coal in 82 years [13]. Meanwhile, the potential of renewable energy is huge (Table 2), and yet it is hardly utilised. Out of the 801 GW of the renewable energy potential, only less than 9 GW or around 1% is utilised to date [14]. Thus, a transition to renewable energy technology is not merely a luxury imposed by the Paris Agreement pledge, it is a necessity; this given the growing national energy demand, its implications on the country's development, and the exhaustion of cheap local fossil fuel supplies. Naturally, a timely utilization of the local renewable energy potential for power generation in Indonesia is a matter of a 1 The Indonesian voluntary pledge is written in the document of Intended Nationally Determined Contribution (INDC) submitted to the United Nations Framework Convention on Climate Change (UNFCCC) in 2015. 2 Calculated from Ref. [9]. 3 Capacity factor is defined as the ratio of an actual electricity generation over a given period of time to the maximum possible electricity generation over the same period of time [7]. 4 Calculated from Statistics Indonesia data, http://www.bps.go.id/. 5 Electrification ratio is defined as the ratio between the number of households with electricity access to the total number of households. 6  careful assessment of costs and prospective benefits. To increase the chances of acceptance and implementation of the Paris target, and to decrease the likelihood of a fossil fuel technology lock-in for the upcoming power system expansion in Indonesia, quantitative scenarios of power system developments that accommodate renewable energy are needed.

Paris agreement and de-carbonization of the power sector
A number of studies discuss de-carbonization strategies in developing countries and the implications in response to their commitment to the Paris Agreement. Grande-Acosta and Islas-Samperio [17] present an alternative scenario for the Mexican power sector, by assessing various mitigation options both on the demand and supply sides. Their study concluded that the alternative scenario assures the Mexican compliance with the Paris Agreement. However, the scenario entails an additional investment of 2 billion US$/year over the analysis period. Likewise, Dalla Longa and van der Zwaan [18] analysed the role of low carbon technologies in achieving Kenya's CO 2 mitigation target under the Paris Agreement. One conclusion of this study is that the deployment of these technologies raises the energy system costs in 2050 ranging from 0.5% to 2% of the country's GDP. Kim et al. [19] investigated the impact of the South Korean's INDC on the power system and electricity market in Korea. The study revealed that the implementation of INDC causes an increase in the electricity price by as much as 8.6 won/kWh. Another study of the South Korean power sector [20] assessed the impact of national policies triggered by the Paris Agreementi.e. renewable portfolio standard and feed-in-tariffon the diffusion of renewable electricity. The study confirmed that the policies indeed influence renewable electricity diffusion. Meanwhile, in response to the Paris Agreement, China needs to radically de-carbonize its electricity sector, which will create unintended consequences, such as the disturbance in stability and integrity due to intermittent renewable energy generation [21]. Thus, Guo et al. [21] present analysis of decarbonization of the Chinese power sector taking into consideration these temporal variations. This study found that the inclusion of temporal variations resulted in a significant difference in terms of installed capacity and load factors when compared to the standard model. Wan et al. [22] assesses the impacts of Paris climate targets on water consumptions of power sector in major emitting economies, which include Brazil, China, India, US, EU and Russia. The study discovered that the fulfilment of long-term climate targets will increase water consumption of power sector, when compared to the business as usual pathways, particularly in the case of China and India.
Studies on de-carbonization of the Indonesian power sector are available in the literature. Marpaung et al. [23] used a decomposition model to examine two factors (i.e. the technological substitution effect and the demand side effect) that affect CO 2 emissions by considering an influence of external costs on the development of the Indonesian power sector during the period of 2006-2025. This study concluded that increasing external cost at a high level allowed for technological substitution, which led to CO 2 emissions reduction by up to 82.5%. Meanwhile, at the low to medium external costs, CO 2 emission reduction was mainly due to demand side effect. Shrestha and Marpaung [24] used an input-output decomposition approach to analyse factors that affect economy-wide change in CO 2 , SO 2 and NOx emissions in the case when the Indonesian power sector employs integrated resource planning (IRP) approach, rather than traditional supply-based electricity planning (TEP). The IRP approach resulted in CO 2 , SO 2 and NOx emission reductions of 431, 1.6 and 1.3 million tons, respectively, during 2006-2025, as compared to that under TEP approach. Rachmatullah et al. [25] used the scenario-planning method to devise a longterm electricity supply plan (1998-2013) of the Java-Bali power system, which included analysis on CO 2 emissions. This study showed that 15% CO 2 emission could be reduced at an abatement cost of around US$2.8-4.0 per ton. Wijaya and Limmeechokchai [26] introduced low carbon society actions into the long-term Indonesian  [12]. [This data include both PLN and IPP productions].  power system expansion planning (2007)(2008)(2009)(2010)(2011)(2012)(2013)(2014)(2015)(2016)(2017)(2018)(2019)(2020)(2021)(2022)(2023)(2024)(2025), using the LEAP model. This study concluded that the low carbon society actions reduced external cost by 2 billion US$ as compared to the conventional electricity expansion planning. Purwanto et al. [27] developed a multi-objective optimization model of long-term electricity generation planning (2011-2050) to assess the economic, environmental, and energy resources adequacy. That study revealed an electricity system scenario with high orientation on environmental protection became the most sustainable scenario, yet lacked in term of Reserve to Production Ratio (R/P) and cost-related indicators. Kumar [28] analysed the effects of different renewable energy policy scenarios on CO 2 emissions reduction, employing the LEAP model. The results showed that utilization of the Indonesian renewable energy potentials reduced up to 81% of CO 2 emissions when compared to the baseline scenario. These studies analysed long-term power system expansions and their associated CO 2 emissions. However, they did not set specific targets on CO 2 mitigation as a constraint in the simulations of future power generation. A few studies set a specific target for CO 2 emission in the Indonesian power sector, and simulated patterns of power supply to meet the targets set. Marpaung et al. [29] developed a general equilibrium model i.e. Asia-Pacific Integrated Model (AIM)/Enduse for Indonesia and analysed the effects of introducing a CO 2 emission targets on the technology mix of the Indonesian power sector during 2013-2035. The study concluded that the deployment of carbon capture and storage (CCS), biomass, and wind power technologies, contributed significantly to achieving the targets. Das and Ahlgren [30] analysed CO 2 reduction targets scenario for the long-term Indonesian power system expansion planning (2004-2030) using the MARKAL model. The results showed that constraints on CO 2 emission invoked changes in technology mixes. Similarly, Stich and Hamacher [31] applied different levels of CO 2 emission reduction targets to the optimized power supply in Indonesia. Their results demonstrated that the CO 2 emission constraints boosted geothermal power expansion to replace the coal-fired power generation. Finally, Siagian et al. [32] used AIM to simulate energy sector development over the period 2005-2030 under CO 2 constraints stipulated in a draft policy document. Their study indicated carbon prices of US$16 and US$63 (2005)/tCO 2 under 17.5% and 32.7% CO 2 reduction scenarios, respectively. Neither of these studies is associated with the prevailing Indonesian policy as stipulated in the final Nationally Determined Contribution (NDC) as submitted to UNFCCC after the Paris Agreement came into force. Our study fills this gap by incorporating the actual energy sector's mitigation targetsi.e. 11% CO 2 reduction on its own effort and 14% CO 2 with international supportinto power capacity expansion. Hence, our analysis can be used as a reference when formulating the legal framework for curbing CO 2 emissions in the power sector.

Review of available models and selection of an appropriate model
Rigorous models for analysing long-term planning are necessary to support any optimal allocation of investments. While numerous energy system models are developed and used worldwide [33], not all are suitable for performing technological, economical, and environmental analysis of the national power sector. Consequently, this calls for the careful selection of a model. Urban et al. [34] provided a comparison of 12 models in terms of their suitability for applications in developing countries. The study argued that the characteristics of an energy system in developing countries differed from those of developed countries. They found that the energy-related characteristics of developing countries, such as supply shortages, urbanization and low electrification rate, were neglected in many models. Bhattacharyya [35] provided an assessment of 12 energy system models suitable for analysing energy, environment, and climate change policies in developing countries. This study suggested that the accounting-type models were more relevant for developing countries because of their ability to capture rural-urban differences, traditional and modern energy technologies as well as non-monetary transactions. Connolly et al. [6] reviewed energy system models to identify ones suitable to analyse the integration of 100% renewable energy into various energy systems. The paper presented a brief summary of 37 models. Most included the power sector in their analysis. However, they varied significantly in their purpose. We filtered the models to identify those that are open access and suitable for analysing the technical, economical, and environmental parameters of a long-term power system expansion in a developing country. The detailed inventory of models and screening criteria are presented in Appendix A (Table A.1).
This screening produced a shortlist of 30 models as being potentially suitable for our analysis (Appendix A, Table A.2). We chose the LEAP model for a number of reasons. Firstly, it can accommodate various characteristics essential for an energy sector analysis in developing countries [34]. Secondly, LEAP is freely accessible to students, as well as non-profit and academic institutions in developing countries. This increases the chances of reproducing and further developing this analysis beyond the efforts in the current article, as and when new data and policy considerations become available. In addition, LEAP is very userfriendly and provides open-access online support for its users. Finally, LEAP is considered as a popular energy model, as it has been used in 190 countries and hosts more than 32,900 members in its online platform [5]. It makes results comparable across countries, which is especially relevant when major international agreements such as the Paris accord are considered.
Indeed, LEAP has been used in many studies for analysing CO 2 mitigation in the power sector worldwide. The LEAP model is employed to explore various scenarios of the energy system development in Taiwan [36], China [37], Iran [38], Panama [39], Maharashtra, India [40], Pakistan [41], and Africa [42]. LEAP has also been actively applied to study the energy and CO 2 impacts of a power system expansion with a special focus on a particular energy source such as landfill gas in Korea [43], nuclear in Korea and Japan [44,45] and renewable in Southeast Asia [28]. None of these studies combined the accounting and optimization settings in LEAP for the analysis. Yet, the features provide for different types of analysis, which is essential for a choice of a robust policy. On the one hand, the accounting method in LEAP can be used to represent future power systems based on various policy scenarios, enabling the comparison of energy, environmental and economic impacts of various energy policies. This method provides answers for a "what if" type of analysis. On the other hand, the optimization method in LEAP optimizes investment decisions and, thus, provides a picture of what could be the optimal pathway of power system development. In this study, we employed both accounting and optimization methods from LEAP to analyse scenarios of CO 2 mitigation in the Java-Bali power system.

Model setup
We started by setting up a LEAP model of the Indonesian power system from the base year of 2005-2015. The structure of the model is shown in Fig. 2. LEAP consists of three modules namely: demand, transformation and resources modules. Electricity generation belongs to the transformation module. The electricity generation module in LEAP simulates electricity supply to satisfy the given demand, based on various input parameters ( Table 3). The model outputs consist of added capacities, a composition of technologies, electricity generation, GHG emissions, and costs.
We initialized LEAP with 2005 data and simulated the expansion of PLN's power generation capacity from 2006 to 2015. According to the PLN statistics, coal steam turbine (CST), natural gas combined cycle (NGCC), natural gas open cycle, diesel generator, hydro, geothermal and solar power were the main technologies employed during this time.
CST and NGCC expanded significantly during this period, i.e. 72% and 18% of the total capacity addition respectively, leaving the remaining 10% of the capacity addition shared between the rest. Thus, the Indonesian LEAP model endogenously simulates CST and NGCC for capacity expansion, while the other technologies are exogenously added. Table 3 lists the type of model outputs, as well as the employed input data and the data used for validation of the Indonesian LEAP model. The input data for electricity demand is the actual values of the national electricity consumption during 2005-2015. The power system characteristics such as transmission losses and a load shape as well as technologies characteristics are based on the actual values. Meanwhile, the data used for validation of the model results consists of historical electricity generation, historical capacity expansion of each technology, and historical fuel consumptions. Appendix B presents detailed information regarding parameters for the Indonesian LEAP model validation.
The simulation of electricity generation in LEAP consists of two steps. First, LEAP calculates the capacity expansion required to satisfy the demand and the capacity reserve of the power system. The outputs of this calculation are the capacity added each year and the composition of technology (capacity mix). Second, LEAP dispatches electricity from each process in accordance with the annual demand and the load curve. The output of the second step is the annual electricity production from each process, which later determines the annual GHG emissions.
As discussed above, LEAP has two different settings for simulating electricity generation: accounting settings and optimization settings (Table 4). We compare the simulation results from both of these settings with the actual data of the year 2005-2015 to validate the model performance.

Validation results
The results show that LEAP calculated the total capacity added (Step 1 in LEAP) during the period of 2006-2015 accurately (Table 5). LEAP slightly underestimated the added capacity, when compared to the empirical data under both the accounting and optimization settings: by just 0.7% and 1.8%, respectively. As a result, in our case the accounting setting was found to be more reliable i.e. 99.3% accuracy, when compared to the 89.2% accuracy of the optimization settings.
The model results also accurately represented the actual technology mix of the capacity added over the period 2006-2015 with 100% and 99% accuracy in accounting and optimization settings, respectively. The optimization setting results, which accurately reproduced the technology mix, indicate that the Indonesian power sector development during this period was based on least-cost principal. This is in line with the PLN's capacity expansion policy as stipulated in the Electricity Supply Business Plan (RUPTL) [46].
The results of total electricity production -Step 2 in LEAPand GHG emissions are shown in Table 6. The calibrated LEAP model calculated precisely the total electricity production in the period between 2006 and 2015, both in the accounting and optimization settings i.e. 100% and 99.93% accuracy, respectively. Meanwhile, the model overestimated the CO 2 emissions by 2.6% and 2.5% in the accounting and optimization settings, respectively.
Based on these results, we concluded that LEAP calculations were accurate. However, as is the case with any other energy model, the simulations in LEAP depend on input data and assumptions. Hence, it may generate uncertainty in outcomes when long run future perspectives are taken.

Future scenarios development
The validation phase (Section 3.2) indicated that the LEAP model estimates of the capacity added, technology mix, cumulative electricity production and CO 2 emissions were reliable. We moved to a scenario analysis of possible future developments of the Java-Bali power system. We developed four scenarios of a capacity expansion for the Java-Bali power system. The scenarios were developed based on the changes in LEAP's assumptions on the choice of a power generation technology and/or a type of energy source. All scenarios aimed to reduce CO 2 emissions in line with the Indonesian pledge to the Paris Agreement.
The first scenario (REF) represents the continuity of the present trend that sets the benchmark. The second and third scenariosnamely: NGS and REN, respectivelyfollow The National Energy Policy 2014 (NEP) [47]. NEP aimed to increase the use of natural gas and new and renewable energy to attain minimum 22% and 23% shares, respectively, in the national energy mix by 2025. The new and renewable energy target in NEP includes nuclear. Nonetheless, NEP emphasizes that nuclear is the least preferable option. These three scenarios use the accounting settings in LEAP, which enable the analysis of different paths of the future of power supply, based on different policy assumptions considering the realistic constrains Indonesia currently faces. Meanwhile, the fourth scenario (OPT) uses the optimization setting of LEAP to find the least-cost solution for the capacity expansion and  Table 3 Input-output parameters in the Indonesian LEAP model and data used for validation. electricity dispatch. As such, this scenario does not account for the realistic policy constrains per se but rather serves as a normative benchmark from the cost minimization perspective.
The objective of all scenarios is to meet the growing electricity demand of the future while fulfilling the Indonesian commitment under the Paris Agreement. According to the Indonesian First Nationally Determined Contribution (NDC) document [48], Indonesia is committed to reduce 29% of GHG emissions by 2030, compared to the business as usual scenario. A more ambitious target is set at 41% reduction, subject to the availability of financial support from developed countries. Given the 29% and 41% targets, 11% and 14%, respectively, should be reduced by the energy sector alone. These targets are lower when compared to those of the forestry sector, which are 17% (own effort) and 23% (with international support). This is expected as the forestry sector 8 contributed 51% of total emissions over the period of 2000-2012, while the energy sector contributed 32% over the same period [49]. With no mitigation policy, the energy sector alone is expected to produce up to 1669 million tCO 2 e of GHG emissions in 2030 [48]. This consists of emissions from the power, transportation and other energy sub-sectors. Yet, the NDC document does not specify the CO 2 emission baseline for the power sector alone. Hence, for all practical purposes, we assume that the power sector contributes proportional to the national energy sector target. Accordingly, we assume that the Java-Bali power system should reduce its emissions at least at the same pace as agreed upon at the national level. Thus, 11% and 14% by 2030 are our targets for setting up the scenarios.
To quantify possible pathways of reaching these competing targets, we specify the four scenarios of development of the Java-Bali power system in the period 2016-2030 as follows: . in the development of renewable energy, which includes geothermal, hydro, biomass, wind and solar PV. In this scenario, the power system expansion maximizes utilization of the renewable energy potential of the Java-Bali islands to achieve the Paris target. This scenario is in line with the NEP 2014, which aims to increase the use of new and renewable energy in the national energy mix. The REN1 scenario aims for the 11% CO 2 reduction target, and REN2 scenario for the 14% CO 2 reduction target. . Optimization scenario (OPT): This scenario uses LEAP's optimization settings to obtain the least-cost solution for the Java-Bali power capacity expansion, while satisfying both the increasing demand and CO 2 reduction targets. The least-cost solution in LEAP is defined as the power system with the lowest total net present value of costs of the system over the entire time horizon. In the OPT scenario, all technologies that are included in the REF, NGS, and REN scenarios are considered for capacity expansion. The OPT1 and OPT2 scenarios assume 11% and 14% CO 2 reduction targets, correspondingly.

Input data
The data source and methodology to calculate the parameters of the Java-Bali power system are provided in Table 7. Rather than relying on the generic LEAP default values, we use national and regional data. Thus, most of the input parameters were collected from governmental reports and the national electricity company, PLN. Since PLN is the sole owner and operator of the Java-Bali power transmission and distribution network, it also records and manages electricity dispatch from Independent Power Producers (IPPs). Table 4 Two alternative simulation settings in LEAP.

Accounting setting Optimization setting
Step 1: Capacity addition LEAP controls when to add new supply capacity based on annual electricity demand.
LEAP controls the type of new supply capacity to be added and when it will be added based on annual demand and cost optimization.
Step 2: Electricity dispatch from each type of power supply LEAP controls the dispatch of electricity from each process based on a user-defined merit order.
Driven by cost optimization, LEAP controls the dispatch of electricity from each process. The demand projection for the year 2016-2025 is calculated based on the estimated demand growth in RUPTL [16]. Meanwhile, the demand projection for the remaining years (2026-2030) was calculated assuming that demand for electricity continues to grow at the same rate as the estimated growth in the year 2026 i.e. 7%. It is assumed that transmission losses will gradually decrease from 8.6% in the year 2015 to 7.9% in the year 2030 [50]. The energy load shape is drawn in the LEAP model based on the historical hourly load data collected from the Java Bali dispatcher unit (P2B), as shown in Appendix C, Fig. C.1 [51]. The planning reserve margin was set at 35% in accordance with the criteria in the National Electricity General Plan 2015-2014 (RUKN) [50].
One limitation in LEAP is that the model does not provide for the expansion of transmission and distribution lines. Hence, this study assumes that there are no constraints in the electricity networks, meaning that electricity supply can be transmitted at any time to any load station. Consequently, calculation of costs in this study does not include costs for transmission lines.
The calculation of electricity generation in LEAP depends on the input of the projected electricity demand in the demand module (Fig. 2). After that, the electricity generation module in LEAP assigns technologies to satisfy the electricity demand. The type of power generation technologies for the capacity expansion includes CST, NGCC, natural gas open cycle, diesel generator, hydroelectric (small and largescale), geothermal, biomass, wind turbine, solar PV and nuclear. The existing coal power plants in Java-Bali use conventional technologies, which have relatively lower efficiencies compared to the supercritical (SC) and ultra-supercritical (USC) technologies. According to RUPTL, to improve efficiency and reduce CO 2 emissions, the future coal power plant's development in the Java-Bali power system would only consider SC and USC technologies [11]. Thus, this study only considered USC coal technology for the new future coal capacity addition. To date, there are no nuclear power plants in Indonesia. However, in the future, development of a large-scale national nuclear energy supply is possible as indicated in NEP. Yet, NEP considers nuclear as the last option after maximizing the use of renewable energy sources.
Technology data of the existing power plants in the Java-Bali system were collected from PLN. These data include the power generating capacity of each technology, planned retirement, heat rate, historical production and capacity factor. The accuracy of the existing power plant data is essential to ensure a reliable base year data, which was used as reference for developing power expansion scenarios. The capacities of existing power plants in the Java-Bali power system are presented in Appendix C, Table C.1.
Technology data for future capacity expansion were retrieved from various studies, as shown in Table 8. The costs characteristic of power generation technologies refers to the cost assumptions of the world energy outlook (WEO) model [52]. Costs data that was not presented in the WEO model 2016 were taken from the Indonesian Energy Outlook (IEO) 2016 [53], which relied on data from the International Energy Agency (IEA) database of the WEO 2015. These parameters were assumed to be constant throughout the entire simulation. Fuel costs of coal and natural gas were taken from the PLN Statistics 2015 [10], while nuclear fuel cost was taken from the EIA study [54] and biomass fuel cost was sourced from ASEAN Energy Centre (ACE) study [55].
The calculation of the resources/fuel requirements in LEAP depends on the outputs of the electricity generation module. LEAP allocates resources needed for the electricity generation in accordance with the fuel efficiency of each technology. It is essential to take into consideration the availability of the energy resources for the Indonesian LEAP model, particularly with regard to renewable energy, as it can only be utilized locally. Hence, the expansion of renewable power generation in this study takes into account the potential of renewable energy in the JB islands. For practical reasons, we assume that the publicly available data for the Indonesian renewable energy potential (Table 1) is accurate and the total renewable potential can be exploited over the time horizon of this study. Furthermore, we assumed that Indonesian natural gas reserves could be utilized for power generation without any constraint.

Reference scenario (REF)
In the REF scenario, the total capacity added during 2016-2030 is 63.8 GW. Thus the power generation capacity in 2030 reaches 94.2 MW (Fig. 3a). The capacity mix in 2030 is equivalent with the base year, when coal, natural gas, and renewables constituted 55%, 33% and 12% of the total capacity, respectively. A small portion of oil capacity (0.3%) **In the accounting setting of LEAP, merit order is the order in which a power plant will be dispatched [5]. * Reserve margin is the percentage of reserve capacity relative to the capacity needed to meet the normal peak demand.
remains present in the capacity mix.
In 2030, the corresponding electricity production from coal increases from 108 TWh in 2015 to 350 TWh. Coal contributes 75% of the electricity production in 2030 while natural gas and renewable energy contribute 18% and 7% respectively (Fig. 3b).
The resulting CO 2 emissions in 2030 are 339 million tCO 2 e or nearly three-fold of the emissions in the base year. These are our baseline for the CO 2 reduction in the mitigation scenarios (NGS, REN, and OPT). Table 9 shows the CO 2 emissions baseline and reduction targets.

Natural gas scenario (NGS)
The results of the LEAP simulations for the NGS scenario indicate that switching from coal to natural gas alone can already deliver both 11% and 14% CO 2 reduction targets. This switching requires around 0.3 billion toe of natural gas, which is equivalent to 8% of the Indonesian natural gas reserve.

NGS1 scenario aiming at the 11% CO 2 reduction
In the NGS1 scenario, the natural gas capacity in 2030 expands up to 45.7 GW; nearly a fivefold increase of capacity over 2015. It is a 15% increase of the gas share in the capacity mix as compared to the REF scenario (Fig. 4a). Consequently, the coal capacity share decreases from 52% in the REF scenario down to 39% in the NGS1 scenario. Compared to the REF scenario, the corresponding electricity production from natural gas increases significantly. In the NGS1 scenario, natural gas constitutes 37% of the electricity generation in 2030; more than twofold compared to its share in the REF scenario. Accordingly, coal power generation decreases from 75% in the REF scenario to 56% in the NGS1 scenario.

NGS2 scenario aiming at the 14% CO 2 reduction
In order to reduce the CO 2 emissions by 14% against the REF scenario, further expansion of the natural gas capacity is required. Our  [5]. a IEA [58]. b OECD/IEA [52]. c DEN [53]. d Calculated based on the ratio of availability of the intermittent plant to the availability of a standard thermal plant [5]. e PLN [10]. f Rothwell and Rust [59]. g IEA and NEA [54]. h Ref. [60]. i ACE [55].  results indicate that it takes an additional 1.8 GW of the natural gas capacity when compared to the NGS1 scenario. Consequently, the share of natural gas power generation in 2030 increases up to 41%, which is 4% higher than the NGS1 scenario (Fig. 4b). Meanwhile, the coal capacity share declines to 38%, leading to a reduced share of the coal power generation of 51% compared to 75% in REF.

Renewable energy scenario (REN)
In the REN scenario, the capacity of renewable energy increases significantly in order to attain both 11% and 14% CO 2 reduction targets. Accordingly, the corresponding electricity production from renewable energy also grows in both REN1 and REN2 scenarios (Fig. 5). The deployment of renewable energy in REN refers to RUPTL, which stipulates the plan to maximize the utilization of hydro and geothermal potentials, and to add up to 5 GW and 2.5 GW of solar PV and wind power capacities respectively by 2025 [16]. Meanwhile, no specific target is set for biomass, although the document mentions the plan to add 0.1 GW of municipal waste power plants. In this study, we assumed that 1 GW and 2 GW of biomass capacities are added in REN1 and REN2 scenarios, respectively, out of the total 7.4 GW biomass potentials of the Java-Bali islands.

REN1 scenario aiming at the 11% CO 2 reduction
The renewable capacity in the REN1 scenario adds up to 21.7 GW or two-fold of the renewable capacity addition present in the REF scenario. Renewables now account for 22% of the total power generation capacity in 2030, split between hydro (9.6%), geothermal (6.8%), wind power (1.5%), biomass (1%), and solar PV (3%). Consequently, the share of renewable electricity generation increases up to 16% of the total power generation in 2030, which is two times higher than in REF.

REN2 scenario aiming at the 14% CO 2 reduction
To accomplish the 14% CO 2 reduction target, 4.8 GW of renewables are added on top of their capacity in REN1. In 2030 the renewables capacity reaches 26.5 GW constituting 26% of the total power generation capacity. The share of each renewable is now slightly increased compared to REN1 i.e. hydro 10.2%, geothermal 6.6%, wind 2.5%, biomass 2% and solar PV 4.9%. Meanwhile, the share of renewable electricity generation adds up to 19% as compared to 7% in REF.

Optimization scenario (OPT)
Based on the simulation results of the OPT scenario, the least cost options for meeting the Paris target is to expand both the renewable energy and natural gas capacities. Accordingly, coal capacity is slightly decreased (Fig. 6).

OPT1 scenario aiming at the 11% CO 2 reduction
In the OPT1 scenario, the renewables capacity addition accounts for 12.8 GW (Fig. 6a). This a significant increase from 7.2 GW in REF and consists of hydro (5.1 GW), geothermal (5.5 GW) and biomass (2.2 GW). Wind, solar PV and nuclear do not appear in this capacity mix, implying that they are less competitive compared to the other technologies. If there are significant changes in, for example, PV technology, it may change in the future. The natural gas capacity increases 8% from 31.3 GW in REF to 33.7 GW in the OPT1 scenario. Meanwhile, the coal capacity decreases 2% from 51.8 GW in REF down to 50.8 GW  in the OPT1 scenario. Thus, the OPT1 cost-efficient electricity generation mix in 2030 consists of coal (66%), natural gas (17%), and renewables (17%). Despite its slightly decreased capacity, the production share of coal remains the highest due to its dominance in 2015 (65% of the total production) and the cheap price of coal resources. This indicates a high chance of lock-in with the coal technology infrastructure, if no other criteria, other than cost minimization, are at play.

OPT2 scenario aiming at the 14% CO 2 reduction
The results of the OPT2 scenario show that 3% supplementary CO 2 reduction can be attained through an addition of 0.5 GW and 1.6 GW of natural gas and biomass capacities respectively as substitutes of coal. These produce an electricity generation mix comprising of coal (63%), renewables (20%), and natural gas (17%), in year 2030 (Fig. 6b).
Biomass capacity in OPT is higher than that in REN because this scenario does not necessarily follow the national power supply plan, which focuses on hydro, geothermal, solar PV and wind. Instead, it opts for the least-cost power supply options from among all available options. This indicates that electricity generation from biomass is cheaper than that from solar PV and wind.
While the renewable capacity in the OPT scenarios are smaller than in REN, the electricity generation from renewables in these scenarios are higher. This is because the renewable technologies in the OPT scenarios (geothermal, hydro and biomass) have relatively higher capacity factors than solar PV and wind power, which constitute the major capacity share in the REN scenario. Meanwhile, the natural gas share in the 2030's electricity generation mix of OPT is lower, i.e. 17% instead of 18% in REF. This is due to the CO 2 constraints set in OPT, which allows more renewable power generation as a substitute for some portions of both coal and natural gas.

CO 2 mitigation costs
In LEAP, the total costs of power system expansion consist of capital costs, fixed operation and, maintenance (OM) costs, variable OM costs, and fuel costs throughout the planning horizon. Table 10 shows the total costs for all our scenarios. The costs of CO 2 mitigation are the difference between the total costs of the Java-Bali power system expansion in the REF and in each mitigation scenario. The total cost of the Java-Bali power system expansion in the REF scenario is equal to 0.1% of the cumulative national GDP during the study period 9 . The model projects increased costs are 1.1%, 1.7% and 2.7% in OPT1, NGS1 and REN1 scenarios, respectively, in order to reduce 11% of the CO 2 emissions compared to REF. Furthermore, reducing 14% of emissions leads to the increased costs of 1.7%, 2.4% and 4% in OPT2, NGS2 and REN2 scenarios, respectively, which can be covered by climate finance as mandated by the Paris Agreement. The CO 2 mitigation costs in each scenario range from 0.002% to 0.004% of the total GDP.
The REN scenarios impose the highest total costs among the three sets of mitigation scenarios, while the OPT scenarios offer the lowest. Hence, the most cost-effective CO 2 mitigation scenario is the OPT scenario, followed by the NGS scenario. The optimization method in the OPT scenario ensures the lowest total costs of the power system expansion in meeting the mitigation target. The OPT covers technology of the REF with an addition of biomass technology. These results suggest that the deployment of biomass has the lowest cost impact in achieving both 11% and 14% CO 2 reduction targets, compared to deploying wind, solar and nuclear technologies. However, the optimal mix can be sensitive to the input assumptions such as the relative performance of different technologies or future costs. Due to the lack of projections regarding changes in costs of different technologies, we keep it aside as  a subject for future research. Table 10 also shows the cost-effectiveness of the CO 2 mitigation scenarios. For the 11% CO 2 reduction target, the abatement cost are 14.9, 22.2 and 30.4 US$/tCO 2 e in OPT1, NGS1 and REN1, respectively. Meanwhile, for the 14% CO 2 reduction target, the abatement cost are 17.7, 29.1 and 41.8 US$/tCO 2 e in OPT2, NGS2 and REN2, respectively. A sensitivity analysis for different discount rates was performed to understand their impacts on cost effectiveness, and particularly for the OPT scenario on technology mix. The results are presented in Appendix D.
Providing reliable and affordable electrical energy for the entire population is a vital development goal of the Indonesian Government. The electricity tariff is determined by the government and subsidies are allocated for low-income households to ensure electrical energy is affordable for all people. In 2017 alone, the electricity subsidy accounts for 3% of the Government's revenue [61,62]. Thus, it is paramount to maintain low electricity production costs. As such, the Indonesian power sector development follows the 'least costs' principal, with electricity supply options chosen based on the lowest cost [11]. Any additional cost, such as the cost incurred due to the compliance with the Paris Agreement, will increase the electricity production costs, which eventually can increase the price paid by consumer and the size of the government subsidy.

Conclusions
Assuring electricity access for the entire population is still a vital national development goal for many developing countries, including the Indonesian Government. Likewise, climate change mitigation and adaptation policies are also essential for developing countries, due to their vulnerability to climate change impacts. Our research analysed plausible trade-offs between electrification and climate goals in the Java-Bali islands. Our focus was on alternatives that could allow Indonesia to satisfy the future electricity demand, while also meeting its Paris climate targets. We chose the LEAP model to perform the analysis after a systematic review of different models (Appendix A). At first, the model was carefully validated using the actual data of the  Overall, our study indicates that, in the contexts of the Java-Bali power system, the Paris target can be met solely by fuel switching from coal to natural gas. Though gas is a cleaner fuel, this strategy would not use the Indonesian potential of its renewable energy. The national electrification goals can also be achieved, without breaking the Paris Agreement, by escalating the development of renewable energy. However, the most cost-effective measure is by increasing renewable energy development in combination with an expansion of natural gas power generation capacity. As far as costs are concerned, our results are aligned with other studies regarding the implication of the Paris Agreement on power sector in developing countries, for example, the case of Korea [19] and Kenya [18].
Our results demonstrate that any effort to comply with the Paris climate target will impact the electricity generation costs. Currently, there is no regulation in place in Indonesia to limit CO 2 emissions from the power sector. Development of such a policy needs to consider the cost implications for the power generation companies, as these costs, eventually, are bound to pass onto the consumers or have to be covered by governmentals subsidies.
We have incorporated the Indonesian Paris targets into the power capacity expansion. However, there is a number of important methodological issues that should be incorporated in any future work. In particular, these types of assessments would benefit greatly by taking into consideration climate change impacts on the future power system, and possible adaptation strategies in the power sector. It is also vital to account for the evolution of energy technologyas most technologies in time become more cost effective [63] and analyse their impact on the robustness of the power system expansion scenarios within the Paris Agreement constraints.   The program is written in GAMS modelling language, which is a commercial software. b Free for students and non-profit institutions in developing countries. c Free for academic purposes and the International Atomic Energy Agency (IAEA) member states.    Table C1.