Operation-driven modeling of depot charging stations for electric bus rapid transit (eBRT) power demand projection for the new capital city of Indonesia

The use of electric mobility (e-mobility) for urban eBRT is expanding globally as more carbon abatement backend technologies are put into practice. Supportive electric power build-up capacity and adequacy readiness infrastructure must be accurately prepared or reconfigured to accelerate the sector’s decarbonization pathway. However, due to the integration of new technologies, which has never been done before, the unavailability of historical data gathered from practical field data or artificially generated data is likely unavoidable. Meanwhile, using as-built data referring to other countries or common presumptions is becoming problematic due to the high possibility of mismatches and uncertainty variables of the diverse scenario, which could lead to oversights and, thus, investment misconduct. Therefore, we construct a model replicating operation-driven eBRT depot charging stations, which focuses on the uncertain domains related to the fleet’s attributes, lane destination, and recharging interval based on the predictive Monte Carlo model (PMCM) to generate time series charging demand required in the early stage of infrastructure expansion within the logical and scientific acceptance. The findings provide insights for all relevant actors, i.e., grid planners, stakeholders, and operators, emphasizing the evidence-based research for climate mitigation action toward Indonesia’s new capital city (INCC) 2045 target.


Introduction
The exponential growth of global e-mobility and its derivative supporting charging-recharging cycle (CRC) instruments and its ecosystem, especially integrated public transport systems (IPTS) envisioning green and integrated transport systems into smart cities concepts, presents substantial critical planning, i.e., strategy, build-up, and scaling as well as market capitalization opportunities for government initiatives while committing climate change strategy.This issue remains one of the most urgent and widely prevalent global concerns, with far-reaching consequences for economies, communities, and the environment.Nevertheless, the pace of net zero emissions (NZE), which seeks to reduce greenhouse gas (GHG), enabling the large-scale abatement up to (below) 1.5 degrees Celsius, necessitates the 2 implementation of innovative and extensive actions, particularly during the initial phases of energy transition (ET) in developing countries that would still heavily rely on fossil fuels for business revenue longevity (in term of negating the oil substitution) for the upcoming [1]- [3].
In light of the preceding trend, deliberate decision-making should be made in the early stages to elevate the prospect of a secure transformation and determine the best course of follow-up action for the future sustainable transport system while also figuring out which infrastructures should be prioritized to get funded on the deployment through potential revenues trajectory and new business models setting up [4] [5].Due to rising activities on the related terms, much attention should be paid to the stakeholders in developing countries, emphasizing infrastructure readiness and asset adequacy level.The notable report estimated that the corresponding grid utility/operator (UO) would spend 1,700 to 5,800 USD per emobility unit for grid modernization by 2030.Approximately 95% of the required budget is devoted to distribution grid assets for investment breakdown expenses.Depending on the case-by-case CRC process, the scenarios represent the cumulative grid upgrading budget of 1.8 billion to 6.2 billion USD over the intended year.Furthermore, the report indicated a clear correlation between the user CRC's preferences and behaviors (which could be exacerbated in terms of site allocation, i.e., public charging services) due to the additional energy cost users should pay and an effort to limit grid modifications, which is a subject that should attract much awareness of the related-actors, i.e., policymakers, operators, investors, and end-users [6][7].
Consequently, the study emphasizes the importance of supportive depot charging infrastructure to the ongoing INCC masterplan of public transport within the principle of 'connected, active and accessible' toward the NZE program by 2045 [8] [9].However, the lack of prior data hinders preliminary assessment planning in capturing the correlation between infrastructure technological diffusion of the eBRT depot and its operation CRC uncertainties, which could hamper the NZE target toward evidencebased actions from technical, economic, and behavioral preparation.By concerning these challenges, to this end, the proposed framework model enables the profiling of power demand consumption by emulating the CRC operation-driven eBRT depot, and thus, optioning the bottleneck issues mentioned above (by filling the adoption gap and readiness impact).It also serves as a timely reminder for UO to quickly adapt and adjust their grids to the ongoing transition, such as IPTS e-mobility and related sustainability initiatives, and simultaneously ensure that they can be feasibly implemented without putting current planning and operations (P&O) and future at risk (infrastructure compatibility, system demand related to planning reserve margin, and real-time hour-to-hour operation affected by the adoption gap and readiness impact).Also, it would be helpful to consider tailored recommendations when considering the prospects of gradual ET toward the long-term sustainability of IPTS programs in terms of the massive adoption of the eBRT depot.This task should be undertaken before selecting and prioritizing the charging infrastructure that supports the most economically efficient IPTS.

Methodology
Figure 1 illustrates the practical eBRT CRC process in part (a) in comparison to the proposed emulating framework model in part (b).The eBRT's depot typically consists of sequential processes of electric supply-demand.Part (a).1 depicts the deliverance process of electrical energy, which is taken as the frontend process, where the supplied distribution grid feeder feeds the depot's main distribution panel (MDP) through a dedicated transformer (DT).Depending on the depot's required capacity (aggregated fleet dispensers and supporting equipment), the point of connection (PoC) might be configured differently based on the contract with the electric supplier (or directly to the UO in the scene where a regulated scenario is established as ongoing in Indonesia).As an overview, the minimum possible connection that aligns with the topic discussion is found within two types: the sixth of the business type (hereafter configured as the third type of business or B3) or the seventh of the industrial class (I3).Both have a minimum 200 kilovolt-ampere (kVA) capacity rating on their DT through medium voltage (MV) PoC within a range of up to 30 megavolt-ampere (MVA) of the required capacity.More details about the supply business license contract for public or private interest, the minimum tariff, and the adjustment rate strengthening the ongoing national electrification transport system can be found in [10] [11].
Generally, the UO serves the end-users with advanced metering instruments to track the power used, the amount of energy used for imposing causes, and related things in a real-time and fair manner as taken in part (a).2.The backend process of the eBRT's CRC operations is reenergizing the depleted internal battery pack (IBP) through the fleet dispenser (FD) as seen in part (a).3.The FD converts alternating current (AC) electric energy fed by the grid (through MDP) into direct current (DC) and stores the energy for driving power.However, since operational-dependent activities and various physical instrument-attributed drive the cycle process, it is hard to avoid many non-linear and dependent factors linked with the occurrence period throughout the CRC process.Due to the complexity issues and the necessity of forecasting the baseline benchmark power demand of the CRC eBRT depot (actuation and aggregated energy usage) for short or long-term planning scenarios, a model framework is presented.
The above CRC processes are emulated into a model that utilizes Google Inc.'s data analytic platform on spreadsheet processor (DAP-SP), which features to exchange data in real-time collectively based on web applications or stand-alone for basic to advanced modeling purposes, and hence, encapsulates the complicated behavior of today's modern power system (MPS), i.e., load and price signal for marketbased and associated operation.Since the DAP has advanced computing features that incorporate artificial intelligence's subset of machine learning (ML) with built-in functions that can fetch data from/any format and platform, a machine can automatically learn from past data (or create new data) without being directly programmed, decomposing complex MPS scenarios within the intended integration (by soft-coupled or coordinated) to a sophistic computer-aided tool such as DIgSILENT Power Factory, etc. Further, the DAP can be exploited into an integrated model system with different data formats or other features, as ref in [12] explains.Next, the brief outline of the remaining model framework is elaborated.

Predictive Monte-Carlo Method (PMCM)
The Monte Carlo Method (MCM) is a numerical technique for solving inquiries in mathematics.Compared to standard random samples, it collects anticipated findings that may be summarized for easyto-understand graphing and tabulation.To accomplish this, the MCM has to generate random numbers (GRNs), a series of numbers that change randomly over time.Prior to creating the GRN, a probability distribution function (PDF), which can be a graph, table, or formula that displays the probabilities connected to each value of the GRN variable, must be derived.Therefore, its derivation is necessary for the related GRN creation.The two distributions most commonly employed in PDF are the normal and uniform distributions [13] [14].With this concept in mind, the uniform distribution function of the CRC In the same manner, the algorithm exploits the boundary function of {minimum, maximum} in each uncertain entity to stochastically generate a set of predicted (within the boundary) conditions as pseudorandom that refer to 'seeded entities', i.e., let us assume the rating IBP capacity (from micro to large size) of eBRT unit is ranging from {75 ⁓ 675} kilowatt-hour (kWh).To acknowledge this, therefore, the uncertainty function (IBP  ) related to IBP capacity, as denoted in (1.b).On the other hand, the word "predictive" emphasizes the function of the seeded boundary, which can be studied further or added to AI's plan for future study.Consecutively, the algorithm executes the eBRT depot cycle operation through the same technique for each different physical attribute and/or operational-driven process.Hereinafter, only three parameters need to be looked at closely: i) the estimated charging cycles ( ⃛ ), ii) the FD plug rating ( ⃛ ), and iii) the estimated charging, stop, and round-trip ( ⃛ ), which are also affected by other unknown composing parameters listed in Table 1 for the intended setup.The sign  ⃛ means that uncertainty elements are directly impacted by other unknown factors, which will be briefly discussed in the next part.

Multi-uncertainties of Operational-driven Parameters
Concerning the perspective of the study goals quantitatively, it is vital to establish a well-functioning classification to break down the multi-uncertainties of operational-driven parameters (MUOP) and make them easier to categorize into the mathematical model formulation.The four categorizations of operational cycles related to various physical attributes and behavioral processes in an arbitrary and unknown manner are derived as follows: i) The uncertainty emerges from the eBRT unit (eBR  ), i.e., the variety of battery capacity ratings and the conversion efficiency of internal board charger, ii) The uncertainty emerges from the fleet dispenser (FD  ) equipment and its premises, i.e., maximum and minimum power charging (PFD), along with its derating factor (DF  ), iii) The uncertainty emerges from the IBP (BT  ) on its state of charge (SoC) daily (minimum and maximum) cycle in maintaining lifespan properties and the storability performance against its manufacturing condition so-called state of health (SoH), and iv) The uncertainty emerges from round-trip lanes (RL  ), i.e., the utilized IBP on its daily trip (SoC/day) and its related round-trip loss/day (TLD  ), thus referring to the estimated round-trip lane (ERL  ) in kilometers per day.For a later setup and approach to the problem formulation, please refer to Table 1.
Meanwhile, stochastic power usage profiling is carried out to address the related issues of the eBRTs' CRC operational-driven.To do so, let us assume that a particle entity of   at the respective direction of time   .By the preposition of the next movement of  +1 , the position is then altered by ∆   in such a manner, the particle is placed into a new position, as denoted  +1 =   + ∆   .Assuming that the changer of ∆   is defined by random events, thus providing an arbitrarily driven continuing process flow.In the same way, the four criteria are written as follows in (2.a -2.d).To better understand the parameters in (2.a -2.d) and their straightforward relation to the actual scheme of the AC charging mechanism [15], as it has been modified, fulfilling the periodic of CRC as depicted in Figure 2. The proposed model framework will then stochastically replicate it based on an evenly distributed random predictive seeding listed in Table 1, aiming for the research objectives.Similarly, the continuous PDF for every FD's sample number FD  is generated with different probabilities on each of its seeds.

Figure 2.
The continuous CRC operation of FD within 1440 minutes timeframe (starting and stopping) plug activation, maximum charging duration T ℎ at the 5 th for 195 minutes, and the idling time T  at the 2 nd for 140 minutes; power charge rating FP   is found ≈ to be 40 kW, modified from [15] Another noteworthy aspect is defining the predictive seed of the parameter idling time T − (T   ) ≈ 5 minutes is defined as a boundary of minimum idling duration (when FDPlug is not in use) due to the minimum delay time needed from CRC session to session to extend the FD's service life.However, in contrast, for T − (T   ) or when the eBRT unit is in active round-trip (in use), it follows the iterative usability scale referring to the CRC operation.

The integrated public transport system (IPTS) of Indonesia New Capital City (INCC)
According to Act No.3 of 2022 concerning the INCC masterplan (detailed in Presidential regulation No.63, which determines the detail of the master plan, and No.64 of 2022, which defines the national strategic area spatial plan by 2022-2042), the coverage area accumulates 256,142 hectares of land and 68,189 hectares of aquatic area.Over the land area allotment, 6,671 hectares are designated as the main area (central administrative governmental area, CAGA), 56,180 hectares (the state capital territory, SCT), and 199,962 hectares as the development area (Nusantara capital region, NCR) [8] [9].
The IPTS, as the intended conceptual design of INCC road infrastructure development, with a focus on enhancing connectivity, activity, and accessibility, has the primary objective to support the development of the road network in Kutai Kartanegara, North Penajam Paser, Samarinda, and Balikpapan, as well as the broader East Kalimantan province.This scenario will be achieved by improving passenger and logistics connections at the regional level, connecting the main center activity in the INCC with the surrounding areas in East Kalimantan province.In response to the demand for road-based IPTS, bus lane connections have been provided to cater to the three city regions and their respective environs.The road development aligns with the notion of the rail-lane corridor, facilitating both intra and interconnections.Additionally, it serves as a means to establish significant freight transport hubs that connect center activity and enhance the integration of transport systems gateways, i.e., airports and seaports.
Given the correlation and approximation of the coverage area in the perspective of intra|inter eBRT feeder (related to the IBP capacity), the quantification distances via route distance are taken on the given zonings and allotments (CAGA, SCT, and NCR) as depicted in Figure 3 based on the Euclidean distance algorithm (EDA) [16].However, the available supporting data is related to road designations, feeder bus  locations, and connecting routes to/from INCC.Meanwhile, the type of eBRT and its related parameters, operating route (start/stop), and the initial route to the final destination have not been found publicly in any data repositories.Therefore, the approximation for the eBRT operational-driven could only be gathered by exploiting the{Min|Max} distance coverage of the planned roads.Consequently, by using EDA, the road can be classified as primary arterial roads, secondary arterial roads, as well as primary and secondary collector lanes which have {11.68 km|408.6 km}, {33.36 km|148.9km}, {56.58 km|172.3km}, {62.78 km|405.2km} respectively.With these in mind, choosing INCC as a study case of eBRT depot CRC operation is aligned with the needs and fulfils the scientific perspectives.

Result and Discussion
This paper proposes a framework model that generates high-resolution data of eBRT depot CRC operation-driven in serving INCC NZE target 2045, which can be used for IPTS infrastructure deployment.Therefore, to achieve economic interests and accelerate the ET in the transport system costeffectively, determining the bottlenecks in the early stage of electric infrastructure deployment instead of using general equivalent estimation will likely lead to significant differences.The CRC eBRT depot operation is covered in formula (2.a-2.d), which is then iteratively computed using DAP-SP to produce high-resolution data of power consumption (per minute) using the predictive seeding parameter setup as listed in Table 1, within a total of   ≈ 83 FD units consist of three plugs per unit.To better overview, the individual CRC plug activation is compared to the typical cycle of AC charging, as seen in Figure 2 and Figure 4.a., which then ensures that the CRC operational-driven of eBRT depot is replicated.Further, Figure 4.b. is considered to validate CRC per day hour-operation (HO), and the aggregated power consumption of a single FD is denoted in the formula (2.c-2.d).Hence, it can be observed that the FP FD  is impacted by DF  and η  marked with peak fall from its power setup FP FD  = 80 kW to {Min|Max} 48 kW-80 kW.Based on both findings, graphs (Figure 4.a.and 4.b.) may have implications for the UO on how uncertain parameters impacted the CRC of FD operation, even in a single activation only.Furthermore, when the IBP  {75 ⁓ 675} kWh have arbitrarily existed, i.e., the depot with a large number and diverse eBRT from the small-to-large size unit; it directly reflects the CRC activation duration (which becomes troublesome due to the average CRC time needed per unit), resulting in accumulative energy used.On that occasion, the depot must have optimal CRC management in matching optimal energy balance operation (technically in feasible scenarios and cost-efficient in minimizing the energy charge) in case the enactment of dynamic tariffs based on time-based or any relevant tariff.
To extensively the findings quantitatively align to the carbon emission abatement and oil displacement while also considering the essential part of infrastructure planning., i.e., distribution feeder transformer, providing improved insight into grid assets management, Figure 4.c.and Figure 4.d.show the accumulative power drawn of a single unit up to the aggregated of 83 FDs unit activations, along with the absolute extremum value (AEV) or the global maximum and minimum of the power curve, which aligns in finding optimal loading (in a reasonable level) when summing the projected load with daily regular load (non-depot load) to preserve the assets' expected lifespan and normal daily operation ≈ 1440 minutes.It can be seen that the accumulative AEV of P − 16.082 MW for 53 minutes and P − 1.111 MW for 4 minutes for 83 FD activations concurrently in serving a total of 1710 units.The energy used is found to be 358.224MWh within the equivalent carbon reduction (CR  ) 37.15 tCO2e and the equivalent oil displacement (OD  ) to be 73.298kiloliters (kLe), as further detailed in Table 3.Nevertheless, the steps of both conversion procedures are as follows: mileage per kilowatt-hour of electricity consumed (kWh)  potential mileage (km per kWh)  carbon emissions as if the mileage were obtained by an ICE (km/kgCO2e)  the possible mileage for oil substitution (km per liter).The data for the configuration came from an in-house study, and some outside sources cited in [17] [18] are listed in Table 2. Notable results are also expressed, which relate to the standard practice in feeder management of knowing the load factor (LF) that represents the most efficient mode of utilizing and operating the electric grid system assets, which is calculated by dividing the average rate by the rate of the highest load curve at a specific time.It demonstrates that the LF value is proportional to the number of CRC activations, even in a slight improvement.Consequently, the highest possible value of LF (close to 1) is, according to Figure 4.d., close to 0.926-0.930when CRC activation is equal to or above 50 FD units.However, assuring the grid capacity assets during minute-to-minute operation is crucial because LF only determines the loading demand accumulated over consecutive HO.In contrast, the minimum LV value

Conclusion
The proposed framework model of operational-driven CRC eBRT depot enables the generation of highresolution data utilizing predictive boundary seeding instead of the averaged or comparable approximation in the absence of preliminary data.The minute-by-minute power profiles (individually or cumulatively as extensible to the number of sampling cycle-activation as needed) include a timeframe, charging duration, incidental power charging, energy used, and the intention of carbon reduction and oil substitution based on the daily distance route.The findings imply that more eBRT FD activations are proportionately aligned with the LF of grid assets at a maximum scale of 0.922⁓0.930above the 40 th unit operation.However, the incidental peak power of CRC is also increased, which increases the exposure risk of grid assets margin capacity under daily normal operation.Therefore, using the gradual results of this study enables UOs to benchmark their existing coverage electric infrastructure in terms of capability, suitability, and readiness whenever no additional physical capacity is preferable or, in case of planning new construction of non-existing, it will insights into the grid planner on how to prepare the long-term utilization (through future HC trajectory) starting from the upstream feeder until the depot PoC and the optimal site allocation.As the number of gradual activations of FD units increases and their influence on the LF scale becomes more pronounced, this could serve as a reference point for HC's optimal integration in the forthcoming days concerning a comparable high-capacity transportation system that prefers minimal physical upgrades to grid infrastructure.Since finding a cost-effective strategy for the INCC's goal of IPTS driven by electric and/or hydrogenpowered vehicles is vital, i.e., to establish short-term or upcoming carbon prices, credit or trading schemes, and taxation structures related to the electrified transport system or further rethink oil revenue reclaim strategies, thus, optimizing P&O of eBRT depot could increase the smooth transition (by minimize the potential risk of failure) in IPTS infrastructure development.Moreover, our outcomes may help INCC stakeholders to decide which supportive grid asset deployment should be prioritized, aligning the carbon abatement of and oil substitution on a significant trade-off toward the NZE's perspective and INCC's 2045 agenda.

Table 1 .
Uncertainty parameters setup for the eBRT framework model

Table 2 .
Parameters set for equivalent carbon-reduction and oil-displacement intention