A Case Study-Based Framework Integrating Simulation, Policy, and Technology for nZEB Retrofits in Taiwan’s Office Buildings

Abstract

Nearly zero-energy buildings (nZEBs) are central to global carbon reduction strategies, and Taiwan is actively promoting their adoption through building energy performance labeling, particularly in the retrofit of existing buildings. Under Taiwan’s nZEB framework, qualification requires both an A⁺ energy performance label and over 50% energy savings from retrofit technologies. This study proposes an integrated assessment framework for retrofitting small- to medium-sized office buildings into nZEBs, incorporating diagnostics, technical evaluation, policy alignment, and resource integration. A case study of a bank branch in Kaohsiung involved on-site energy monitoring and EnergyPlus V22.2 simulations to calibrate and assess the retrofit impacts. Lighting improvements and two HVAC scenarios—upgrading the existing fan coil unit (FCU) system and adopting a completely new variable refrigerant flow (VRF) system—were evaluated. The FCU and VRF scenarios reduced the energy use intensity from 141.3 to 82.9 and 72.9 kWh/m²·yr, respectively. Combined with rooftop photovoltaics and green power procurement, both scenarios met Taiwan’s nZEB criteria. The proposed framework demonstrates practical and scalable strategies for decarbonizing existing office buildings, supporting Taiwan’s 2050 net-zero target.

1. Introduction

Energy efficiency has long been regarded as the “first fuel” of the clean energy transition, representing one of the most cost-effective and rapidly deployable strategies for achieving global net-zero emissions goals. As a major contributor to energy consumption and carbon emissions, the building sector has become a central target for decarbonization. Within this sector, existing buildings—many of which were constructed before energy performance standards were in place—represent both a challenge and an opportunity. Recognizing this, the European Union updated the Energy Performance of Buildings Directive (EPBD) in 2024, mandating that existing buildings should be transformed into zero-emission buildings by 2050 [1]. Taiwan has also outlined a national roadmap to reduce building energy use by 50% by 2050 compared to 2000’s levels. By 2040, 50% of existing buildings are expected to be retrofitted to near-zero energy levels, with this target increasing to over 85% by 2050. To support this transition, the government has introduced the Building Energy-Efficiency Rating System (BERS) [2], which aims to incentivize voluntary upgrades and facilitate the adoption of nearly zero-energy buildings (nZEBs).

The concept of nZEBs has thus emerged as a key strategy in building-sector decarbonization. Although a universally accepted definition or technical standard for nZEBs is still lacking, most countries adopt context-specific interpretations, treating nZEB as a flexible framework rooted in shared principles rather than a fixed international norm [3]. Common features include ultra-low operational energy demand, highly efficient building envelopes and systems, significant use of renewable energy, and performance evaluation based on site or source energy. Within the European Union, the EPBD provides a regulatory framework for nZEB calculation, referencing ISO 52000-1 [4] and EN 15603 [5] standards, which define the scope of energy services assessed—specifically heating, cooling, ventilation, and domestic hot water—to ensure consistency across evaluations. Specifically, under Taiwan’s nZEB framework, buildings must achieve an A⁺ energy performance label, with more than 50% of the required energy savings derived from retrofit technologies. Despite these policies, practical nZEB retrofit cases in Taiwan’s office sector remain limited, and building owners often lack integrated guidance connecting technical measures with policy benchmarks.

However, despite these policy efforts, nZEB retrofit cases in Taiwan’s office sector remain limited. One of the key obstacles is the lack of motivation and operational clarity among private building owners, who often struggle to access integrated guidance that links technical options with policy benchmarks. In this context, the financial sector plays a growing role in enabling green transformation. In Taiwan, real estate financing has long been a core activity for banks, and integrating energy efficiency or green building standards into lending assessments can help advance green finance innovation. By doing so, financial institutions could guide capital toward sustainable practices, increasing decarbonization awareness in the building industry and accelerating progress toward net-zero goals.

For example, the United Kingdom has successfully leveraged its Energy Performance Certificate (EPC) database as a transparent market indicator, aligning policy and finance to encourage building upgrades [6]. This study responds to a similar need in Taiwan by proposing a locally adapted retrofit evaluation framework, focused on small- to medium-sized commercial buildings. Supported by Taiwan’s national nZEB demonstration program, we adopt a representative bank office building in Kaohsiung as a case study, in part because of the symbolic and practical relevance of banks to the green finance ecosystem.

The proposed framework integrates on-site monitoring, simulation-based analysis, and performance benchmarking under Taiwan’s nZEB criteria. Emphasis is placed on feasible and replicable energy conservation measures and renewable energy strategies, resulting in a practical and scalable workflow to support the decarbonization of existing buildings. Furthermore, by examining the retrofit process of a financial-sector building, this study also addresses the growing intersection between green finance and technical evaluation, providing an operational bridge among policy, market, and industry stakeholders. This framework not only enhances local applicability and implementation potential but also offers a valuable reference for Taiwan’s 2050 net-zero building roadmap.

2. Literature Review

Nearly zero-energy buildings have been widely explored as a strategic solution to reduce emissions and energy demands in the building sector, particularly through retrofitting existing office buildings. Numerous studies have investigated energy-saving potentials and retrofit strategies using simulation models, empirical validations, and economic analysis. A common theme across these works is the necessity of integrating passive design, high-efficiency systems, and renewable energy sources to meet nZEB goals.

For instance, Ballarini et al. [7] demonstrated that upgrading building envelopes, implementing shading devices, and integrating heat pumps with solar thermal technologies significantly improved the energy performance of Italian office buildings. Similarly, Ferrari and Beccali [8] showed that customized technology packages can reduce primary energy demand and CO₂ emissions by over 40%. Simulation-based studies in Saudi Arabia [9] and Brazil [10] reported that combining envelope improvements with HVAC and lighting upgrades yielded energy savings of 25–46%.

Nevertheless, several studies note that energy efficiency measures alone are insufficient to achieve full nZEB performance. Luddeni et al. [11] highlighted, via national simulations in Italy, that photovoltaic (PV) integration alone could not meet nZEB targets without robust policy support. In the UAE, Alkhateeb and Abu-Hijleh [12] found that passive strategies contributed only 14.7% energy savings, while active systems contributed 63.2%. Full net-zero electricity was only achieved when PV systems were included.

Performance gaps between simulation and real-world outcomes are also frequently discussed. In Texas, Shin et al. [13] found that a retrofitted office building with high-efficiency HVAC and PV systems achieved 37–50% energy savings, with PV generation exceeding annual electricity consumption. Conversely, Zhou et al. [14] reported that a project in Tianjin exceeded its expected energy use post-occupancy due to overestimated modeling assumptions, reduced PV efficiency caused by pollution, and inadequate operations—highlighting the importance of design–operation integration.

Climate- and region-specific factors further influence retrofit strategies. Hamza et al. [15] developed a retrofit framework for 1970s–1980s buildings in Egypt using mixed-mode ventilation and PV integration, achieving 46–65% simulated energy savings. In Singapore, Sun et al. [16] combined solar chimneys with efficient cooling technologies to suit tropical conditions, though high costs and immature technologies remain limiting factors. Studies in Pakistan [17] and Kuwait [18] emphasized the value of local knowledge and adaptive strategies in achieving nZEB under constrained conditions.

Economic feasibility remains a significant concern. High upfront costs and long payback periods deter private owners. Luddeni et al. [11] and Zhou et al. [14] both cited lack of incentives as a key barrier to widespread adoption. However, Constantinides et al. [19] showed in Cyprus that with government subsidies and efficient system integration, buildings were upgraded from energy class C to A, with a 92% reduction in carbon emissions.

Some studies argue that meaningful energy savings can be achieved without major equipment replacement. AbuGrain [3] demonstrated that optimizing HVAC and lighting operations alone reduced energy use by nearly 30%, while a modest PV system covered the remaining demand. In Denmark, Rose and Engelund [20] found that building envelope and system upgrades in cold climates reduced energy use to 41.2 kWh/m²·yr—an 84% decrease.

Recent studies have increasingly emphasized the importance of post-occupancy evaluation (POE) in validating building energy performance and narrowing the gap between design assumptions and actual operational outcomes. Menezes et al. [21] demonstrated that discrepancies in energy use often stem from unrealistic modeling assumptions regarding occupant behavior and facility management. By integrating POE data into simulation models, they were able to reduce prediction errors to within 3%. Carletti et al. [22] reinforced the significance of POE by documenting deviations between predicted and measured performance in a monitored nZEB residential building. Cozza et al. [23] further argued that the lack of systematic post-retrofit monitoring in Switzerland undermines policy effectiveness, as performance gaps tend to be more pronounced in high-efficiency buildings. To address this, they proposed the adoption of tools such as the Building Renovation Passport to support long-term performance tracking and policy feedback. Building upon these insights, Zhao et al. [24] and Elsayed et al. [25] conducted comprehensive reviews of POE practices, highlighting the role of occupant-centric feedback in achieving energy-efficient and comfortable buildings. Their findings suggest that incorporating thermal comfort preferences, behavioral data, and user–system interaction into design and operational strategies can promote more adaptive, sustainable building environments. In parallel, energy performance gap (EPG) studies have provided further insights into the persistent mismatch between simulated and real-world energy performance. Zheng et al. [26] identified key drivers of EPG—including modeling errors, occupant behavior, and system control limitations—across various building types. Wang et al. [27] reported dynamic EPGs ranging from 3.0% to 53.5% in five Chinese office buildings, with HVAC control emerging as a major source of inefficiency. Geng et al. [28] proposed a simplified diagnostic method using utility bills and weather data to identify operational inefficiencies and guide targeted retrofits. Similarly, Juričić et al. [29] found substantial performance gaps in Croatian buildings, stressing the importance of performance tracking and longitudinal energy audits in designing effective retrofit strategies. Collectively, POE and EPG research underscore the critical role of empirical feedback, realistic modeling, and long-term monitoring in improving energy prediction accuracy, supporting evidence-based renovation, and informing policy decisions.

Together, these studies reveal a range of effective strategies for nZEB retrofitting, while also underscoring the importance of climate context, integrated design and operation, and strong policy support. Despite progress in other countries, similar localized research in Taiwan—especially in the small- to medium-sized office sector—is still scarce, highlighting the need for a comprehensive, context-specific retrofit framework.

3. Materials and Methods

3.1. Core Principles of nZEB and Taiwan’s Building Energy Rating System

Taiwan has followed similar international developments by introducing the BERS in 2022, informed by the EU’s EPBD experience. The BERS establishes a seven-level energy performance grading system based on energy use intensity (EUI), referencing ISO 52003-1 [30] as a foundational model. Under this scheme, a building is recognized as meeting nZEB status if it achieves at least a 50% reduction in energy use through high-efficiency design measures and incorporates renewable energy to attain an A⁺ Grade.

In 2024, Taiwan expanded this framework by launching a formal energy labeling mechanism targeting existing non-residential buildings. The labeling scheme evaluates operational energy performance based on four key systems: air-conditioning, lighting, elevators, and domestic hot water. Only systems that are present in the building are assessed, with elevators and hot water excluded if not installed. The energy usage of these systems is compiled from utility bills covering a two- to four-year period and is normalized to calculate the building’s EUI. This value, after accounting for onsite renewable generation, serves as the basis for the assigned energy rating.

Table 1. EUI Classification for Small- and Medium-sized Office Buildings in Taiwan.

Grade	A+	A	B	C	D	E	F	G
EUI (kWh/m2·yr)	≤52	52–60	60–68	68–76	76–83	83–108	108–159	>159

presents the EUI classification thresholds for small- and medium-sized office buildings under Taiwan’s BERS. An A⁺ grade requires an annual EUI of 52 kWh/m²·yr or less, while buildings with EUI above 159 kWh/m²·yr fall into the G category.

While the use of utility-derived EUI offers practical advantages in assessing real-world building performance and monitoring policy outcomes, its limitations must be acknowledged. The EUI, being a composite metric, does not isolate individual system inefficiencies nor provide specific retrofit guidance. This may reduce its effectiveness in identifying targeted performance upgrades for underperforming buildings, underscoring the need for supplementary technical assessment frameworks to support deep energy renovations.

3.2. Case Building for Retrofit

To explore practical pathways for transforming existing low-performance office buildings into nZEBs, this study adopts a real-world demonstration approach. The selected case involves a privately owned, small-to-medium-sized office building that remained in continuous use throughout the retrofit process. Given the absence of government subsidies, the project prioritized a site with a high potential for energy improvement, a strong corporate social responsibility commitment, and a willingness to collaborate throughout the implementation phase.

In consideration of the demonstration value and capital-guiding potential of the financial sector in promoting green buildings, we chose a bank branch from among several willing private-sector candidates. The selected branch is of a scale typical to many bank offices in Taiwan, making it highly representative.

The case site is a bank branch occupying the first and second floors of a six-story commercial office building located in downtown Kaohsiung City, southern Taiwan. The total floor area is 630.82 m², with the building oriented 20° east of true north. It features a concrete wall envelope and windows fitted with low-emissivity glazing (solar heat gain coefficient, SHGC = 0.5), and a window-to-wall ratio of 0.16. Internal loads primarily arise from lighting, office equipment, and occupant activities.

The building operates from 8:00 AM to 6:00 PM on weekdays, with energy use concentrated in lighting, HVAC, and plug loads. Prior to the retrofit, the lighting system consisted of 20 W and 40 W T8 fluorescent tubes with a total installed capacity of 7.88 kW. The HVAC system, which had been in operation for over 11 years, included rooftop scroll chillers, a cooling tower, water pumps, fan coil units (FCUs), and two dedicated outdoor air systems (DOAS). In this study, plug loads are categorized into three groups: (1) individual workstations, such as desktop computers and monitors, which are generally used continuously during office hours; (2) business-related equipment, including printers, ATMs, and LED signage, which may operate in standby mode or continuously—with devices like ATMs and signage running 24/7; and (3) breakroom appliances, such as water dispensers, refrigerators, coffee machines, and rice cookers, which are used intermittently throughout the day, reflecting typical staff routines. Although the simulation employed typical-day data with fixed load profiles, these assumptions were informed by on-site observations to capture realistic operational patterns. The plug load was measured at approximately 16 W/m². Annual electricity consumption in 2023 totaled 112,886 kWh, ranging monthly from 7434 kWh (January) to 11,172 kWh (August). Specifications for the pre-retrofit HVAC system are provided in

Table 2. Specifications of Pre-retrofit HVAC System.

Equipment	Specification
Chiller	Two scroll chillers, total cooling capacity: 60 RT; total input power: 51.2 kW
Cooling Tower	Two towers, total heat rejection capacity: 80 RT; input power: 1.5 kW (no VFD)
Cooling Water Pump	Two pumps, total flow rate: 630 LPM; total input: 7.5 kW
Chilled Water Pump	Two pumps, total flow rate: 504 LPM; total input: 7.5 kW
Fan Coil Units	21 units; airflow: 680–1360 CMH; total power: 2.2 kW
Outdoor Air Units	Two DOAS units; airflow: 1600 CMH; input: 1.16 kW (no heat recovery)

.

In addition to conventional energy efficiency upgrades, a 51.2 m² rooftop PV system was installed on-site. Electricity generated is primarily consumed within the building, with excess production stored in batteries for weekday use. To further offset carbon emissions, the bank also participates in a green power wheeling program. From December to April, approximately 28,388 kWh of renewable electricity—equivalent to 25.1% of the building’s annual electricity consumption—is procured from off-site solar or wind sources through the national grid, enabling the site to support clean energy transitions despite limited rooftop space.

The building is situated at coordinates 22.636° N, 120.293° E, in a hot and humid climate zone typical of southern Taiwan. Based on typical meteorological year (TMY) data [31], the region experiences a maximum temperature of 35.4 °C, a minimum of 8.7 °C, and an average annual temperature of 24.8 °C, with average relative humidity at 79.1%. Figure 1 presents monthly cooling degree hours (CDH), calculated with a 23 °C base temperature, along with monthly global horizontal solar radiation. CDH peaks at 4569 °C·h in August, indicating high annual cooling demand. Solar radiation exceeds 4000 Wh/m² from February to September, peaking at 5441 Wh/m² in June, demonstrating strong potential for solar energy utilization.

3.3. Retrofit Procedure

As outlined in Section 3.1, Taiwan’s BERS defines the scope of nZEB performance assessments to include energy consumption from air conditioning, lighting, and elevator systems. Since the case building analyzed in this study does not include elevators, only the air conditioning and lighting systems were considered in determining compliance with nZEB criteria. The renewable energy supply was derived from both an on-site PV system and externally procured green electricity. Whether the building satisfies the nZEB standard is determined based on the energy savings achieved from air conditioning and lighting systems, and the degree to which those loads are offset by renewable sources.

The retrofit process encountered several real-world constraints, including the requirement for continuous building operation without service interruptions, high initial capital investment, varying levels of stakeholder acceptance, physical and structural limitations of the existing building, challenges in technical integration, spatial constraints for renewable installations, and limited flexibility in construction scheduling. To address these challenges, a systematic and collaborative retrofit strategy was developed by a professional engineering team in close consultation with the building owner and interdisciplinary experts.

The retrofit strategy was grounded in the integration of site-specific characteristics, existing mechanical and electrical system configurations, and owner decision-making priorities. Special emphasis was placed on cross-system coordination and adaptive planning to ensure that the selected measures were technically feasible, economically viable, and practically implementable under real-world conditions. This approach also aimed to enhance the replicability and scalability of the retrofit model for other small- and medium-sized commercial office buildings in Taiwan.

The study followed a multi-stage methodological framework:

• Baseline Assessment: Onsite measurements and analysis of historical utility data were conducted to establish an operational energy baseline, focusing on air conditioning, lighting, plug loads, and thermal envelope performance.
• Gap Analysis via BERS: The existing performance was evaluated using the BERS framework to identify gaps relative to nZEB benchmarks. This assessment informed the development of retrofit priorities.
• Measure Development: Targeted energy-saving strategies were proposed through iterative consultations with stakeholders and domain experts, balancing performance gains with practical feasibility.
• Feasibility Evaluation: Each proposed measure was re-evaluated using the BERS to estimate potential energy savings and assess its contribution toward meeting nZEB criteria.
• Simulation and Validation: A detailed energy simulation model was constructed in EnergyPlus to compare pre- and post-retrofit performance. The model was calibrated against actual measurement data to ensure accuracy and reliability.

By integrating field data, standardized evaluation methods, stakeholder engagement, and calibrated simulation modeling, this study established a structured and evidence-based retrofit workflow, as illustrated in Figure 2. The proposed procedure provides both a technical roadmap and a practical decision-making reference for advancing the low-carbon transition of small- and medium-sized office buildings toward Taiwan’s 2050 net-zero targets.

3.4. Selection of Appropriate Energy-Saving Measures

Following the systematic retrofit framework outlined in Section 3.3, this section presents the selection rationale and technical specifications of the adopted energy-saving measures, with an emphasis on lighting and air conditioning systems. Figure 3 illustrates the exterior configuration of the case building. The bank branch occupies the ground floor and the second floor of a multi-story commercial building. Among the four façades, only the front façade is directly exposed to outdoor conditions, while the upper floors of the building are recessed, providing effective self-shading for the occupied levels below. This architectural configuration contributes to passive thermal control by reducing solar heat gain on the building envelope.

The branch is equipped with low-emissivity (Low-E) double glazing (solar heat gain coefficient, SHGC = 0.56), supplemented by interior blinds to mitigate glare and further limit solar radiation. An automatic sliding door is installed at the main entrance on the ground floor to prevent conditioned air from escaping. The rear and side walls are shared with adjacent buildings and function as interior partitions, thereby minimizing thermal losses to the exterior environment.

Based on findings from on-site surveys, performance diagnostics, and discussions with the building owner regarding operational priorities and retrofit feasibility, it was determined that the existing building envelope already incorporates several passive energy-saving features. Therefore, retrofit efforts were concentrated on upgrading the lighting and air conditioning systems, which were identified as the primary contributors to the building’s operational energy consumption.

The original lighting system consisted of fixtures without any verified energy efficiency certification, and field measurements indicated instances of excessive illuminance, leading to unnecessary energy consumption and suboptimal visual comfort. The retrofit intervention involved a complete replacement of all existing lamps with certified energy-efficient lighting products. Furthermore, the number of four-tube fluorescent fixtures was reduced from 169 to 148 sets based on revised lighting layout optimization, while the 30 existing single-tube fixtures were retained. As a result, the total lighting power demand was reduced significantly—from 7.88 kW to 4.10 kW, representing an approximate 48% decrease in lighting power density. The detailed layout of the improved lighting configuration is shown in Figures S1 and S2.

For the air conditioning system, two retrofit options were proposed. The first, referred to as the FCU scheme, entailed upgrading the performance of the existing FCU system while retaining key components. The second, termed the VRF scheme, involved replacing the entire air conditioning system with a VRF system, which is widely recognized for its high energy efficiency in small- to medium-sized commercial buildings.

After careful consideration of capital investment, implementation complexity, and the need to minimize operational disruption, the building owner selected the FCU scheme for actual implementation. Nonetheless, to provide a broader perspective and assess relative performance, both the FCU and VRF schemes were subjected to detailed simulation and comparative analysis. The technical specifications for both retrofit options are summarized in

Table 3. Specifications of Air Conditioning Equipment—FCU Scheme.

Equipment	Post-Retrofit Specification
Chiller	One screw chiller, cooling capacity: 50 RT, input power: 34.1 kW, with variable frequency drive (VFD)
Cooling Tower	Two cooling towers, heat rejection capacity: 80 RT, each with 1.5 kW VFD fan
Cooling Water Pump	One pump, flow rate: 650 LPM, input power: 3.7 kW
Chilled Water Pump	One pump, flow rate: 500 LPM, total input power: 3.7 kW
Fan Coil Units	Retained from pre-retrofit system
Outdoor Air Units	Two DOAS (Dedicated Outdoor Air Systems), airflow: 1600 CMH each, input power: 1.16 kW each, with heat recovery and CO2-based demand control ventilation (DCV)

(FCU scheme) and

Table 4. Specifications of Air Conditioning Equipment—VRF Scheme.

Equipment Name	Quantity	Capacity	Input Power
Outdoor Unit (air-cooled)	1	78.5 kW	25.97 kW
	1	90 kW	28.9 kW
Indoor Unit (ducted)	2	570 CMH	59 W
	20	1500 CMH	184 W

(VRF scheme). For more detailed information on the improved FCU and VRF systems, please refer to Figures S3 and S4.

3.5. Building Energy Simulation of Retrofit Scenarios

To evaluate the energy-saving potential of the proposed Energy Efficiency Measures (EEMs), a calibrated building energy simulation model was developed as part of this study. The model incorporated a comprehensive set of input parameters, including building geometry and orientation, envelope thermal properties, window specifications, lighting and plug load densities, HVAC system configurations, interior zoning layout, occupant schedules, and localized climate data. Critical operational parameters—such as plug load intensity, lighting schedules, and HVAC thermostat setpoints—were calibrated based on on-site surveys, architectural documentation, and consultations with facility management personnel to ensure alignment with actual usage patterns.

A particular consideration in this modeling process was the aging condition of the HVAC equipment. The case building’s chilled water system had been in operation for over 11 years, and its actual performance had significantly deteriorated compared to its original design specifications. Directly applying nameplate efficiency ratings in the simulation would have introduced considerable deviation from real-world performance, thereby compromising the credibility of the simulation results.

To enhance the accuracy of the model calibration and to better reflect current operating conditions, a targeted one-week field monitoring campaign was conducted. This included the collection of real-time data on power consumption for key HVAC components—namely chillers, cooling water pumps, chilled water pumps, and cooling tower fans—as well as chilled water flow rates and inlet/outlet temperatures. Measurement instruments included electrical power analyzers for equipment loads and ultrasonic or thermal energy meters for flow and temperature data.

The simulation was conducted using EnergyPlus version 23.2, with localized typical meteorological year (TMY) weather files representing the regional climate. Model calibration involved iterative adjustments to input assumptions and system parameters, comparing simulated monthly electricity consumption with actual utility billing data recorded in 2023. The process continued until the model outputs fell within industry-accepted tolerances.

To quantitatively assess the calibration accuracy, two statistical indicators were employed: the normalized mean bias rrror (NMBE) and the coefficient of variation of the root mean square error (CV(RMSE)). According to the calibration criteria specified in ASHRAE Guideline 14 [32], a model is acceptably calibrated when NMBE falls within ±5% and CV(RMSE) remains below 15%.

The equations for calculating NMBE and CV(RMSE) are expressed as follows:

$$\mathrm{N}\mathrm{M}\mathrm{B}\mathrm{E}={\displaystyle \frac{\sum _{i=1}^n{S}_i-M_i}{(n-1)\times M_{avg}}}\times 100$$

(1)

$$\mathrm{C}\mathrm{V}(\mathrm{R}\mathrm{M}\mathrm{S}\mathrm{E})={\displaystyle \frac{\sqrt{{\sum _{i=1}^n{(S}_i-M_i)}^2/(n-1)}}{M_{avg}}}\times 100$$

(2)

where:

• $n$ is the number of data points;
• $S_i$ and $M_i$ are the simulated and measured energy use, respectively;
• $M_{avg}$ is the average measured energy use.

The successfully calibrated model served as a robust analytical platform for evaluating the energy performance of the two HVAC retrofit schemes—FCU and VRF—described in Section 3.4. The simulation results were used to quantify annual energy use, estimate energy savings, and assess compliance with nZEB targets under each scenario.

3.6. Renewable Energy Generation via Photovoltaic System

To complement the energy-saving measures and assess the building’s capacity to meet nZEB criteria, a rooftop PV system was incorporated into the energy modeling framework. The installed PV system is comprised of multicrystalline silicon PV modules, each rated at 300 Wp, with other specifications shown in

Table 5. Main technical parameters of PV modules.

Parameter	Unit	Numerical Value
Geometry	mm	1640 × 992 × 40
Peak power	Wp	300
Open circuit voltage	V	39.24
Short circuit current	A	9.93
Operating voltage	V	31.51
Working current	A	9.52
Number of cells on a PV module	Piece	60

. The total module area is 51.2 m², with panels oriented due south and mounted at a fixed tilt angle of 22°, which is optimized for annual solar irradiance in the local climatic context.

PV generation is inherently influenced by multiple environmental and system-level variables, including module efficiency, azimuth and tilt angle, partial shading from nearby structures, and local weather conditions. To capture these influences accurately, this study employed the equivalent single-diode model embedded within EnergyPlus, which is adapted from the TRNSYS PV performance model [33]. This physics-based modeling approach has been widely validated in academic literature for its effectiveness in simulating the electrical behavior of PV arrays under real-world operating conditions [13,34,35,36,37,38,39].

The single-diode model represents the PV cell as an equivalent circuit comprising a photo-generated current source in parallel with a diode and series/parallel resistances to account for internal electrical losses. Input parameters required for this model—such as temperature coefficients, reference efficiency, short-circuit current, and open-circuit voltage—were derived from manufacturer datasheets for the selected PV module.

Shading analysis was also incorporated using site-specific geometry data, enabling the model to account for potential obstructions from adjacent buildings throughout the year. As a result, the simulation provided a realistic estimate of annual PV electricity generation, which was subsequently used to evaluate the extent to which renewable energy can offset the building’s operational loads.

4. Results and Discussion

4.1. In Situ Performance Monitoring and Simulation Model Validation

Given that the existing chilled water HVAC system had been in operation for over 11 years, its performance had noticeably deteriorated compared to the original design specifications. To capture the system’s actual operating behavior, a targeted five-day on-site monitoring campaign was conducted from 20 May to 24 May 2024. The measured electricity consumption, cooling load, and system efficiency during this period are illustrated in Figure 4, which presents both the total chiller power use and the cooling performance across the monitoring timeframe. During this monitoring period, the total electricity consumption of the chiller system was measured at 1215 kWh, with daily consumption ranging from 225 kWh to 262 kWh. The load profile exhibited consistent daily demand, suggesting stable building usage patterns under typical late-spring cooling conditions. An energy end-use breakdown indicated that the chiller compressor constituted the largest share of consumption at approximately 69%, followed by the cooling water pump (15%), chilled water pump (13%), and the cooling tower fan (3%). These proportions are representative of conventional water-cooled chilled water systems operating at partial load conditions. The total cooling load over the monitoring period was 697 refrigeration ton hours (RTh), with a daily average of 139 RTh. The unit cooling efficiency of the chiller—expressed in kW per RT (i.e., specific electricity input per unit cooling output)—ranged from 1.10 to 1.25 kW/RT, averaging 1.21 kW/RT. When the energy consumption of auxiliary systems (pumps and cooling tower fan) was included, the overall system-specific cooling energy input ranged from 1.67 to 1.80 kW/RT, with an average of 1.74 kW/RT.

These measured values served as key reference data for model calibration, ensuring that the simulation accurately reflected the degraded real-world performance of the aging HVAC system. Incorporating these field observations into the EnergyPlus model substantially improved its predictive accuracy and reinforced the credibility of the subsequent scenario analyses presented in this study.

Following the integration of field measurements into the calibration process, the validated building energy simulation model was employed to assess the baseline energy performance of the case building and to evaluate potential retrofit scenarios. The simulation was conducted using EnergyPlus, incorporating detailed geometric and construction information for each floor, including floor slabs, exterior walls, interior partitions, and fenestration systems. These elements were modeled based on as-built architectural drawings to ensure physical accuracy. Operational parameters for HVAC systems, lighting, and plug loads were derived from on-site surveys, while the measured chiller system efficiency, obtained from the aforementioned field monitoring, was directly inputted into the model to reflect actual equipment performance. This hybrid approach, combining empirical data with detailed physical modeling, contributed to the high fidelity of the simulation results. Climatic conditions were represented using the Typical Meteorological Year (TMY) dataset, which is commonly adopted in building energy performance assessments in lieu of real-time weather data. Since the case building lacks a sub-metering system, the calibration process was conducted using total monthly electricity consumption data.

As shown in Figure 5, the simulated monthly electricity use closely aligns with measured utility data from 2023. The model predicted an annual electricity consumption of 111,785 kWh, equivalent to an EUI of 177 kWh/m²·yr, compared to an actual consumption of 112,886 kWh, or an EUI of 179 kWh/m²·yr, representing an accuracy of approximately 99%. Monthly deviations ranged between –15% and +10%, attributable primarily to fluctuations in ambient conditions, occupancy behavior, and plug load variability.

Following iterative calibration, the final model achieved a normalized mean bias error (NMBE) of 1% and a coefficient of variation of the root mean square error (CV(RMSE)) of 7.5%, both well within the thresholds stipulated by ASHRAE Guideline 14 (NMBE ≤ 5%, CV(RMSE) ≤ 15%). These results confirm the robustness and reliability of the model for subsequent energy scenario evaluations. Based on the calibrated baseline model, the pre-retrofit annual electricity consumption breakdown for the bank branch was as follows:

• HVAC systems: 72,737 kWh (65% of total).
• Lighting systems: 16,438 kWh (15%).
• Plug loads: 22,611 kWh (20%).

The combined energy use intensity (EUI) for HVAC and lighting was calculated at 126.32 kWh/m²·yr, which corresponds to a Grade F under Taiwan’s BERS (see Table 1). This indicates a relatively poor baseline energy performance and underscores the necessity of targeted retrofit interventions.

4.2. Energy Performance of Retrofit Schemes

This study employed energy simulation analyses to evaluate the performance of two proposed building retrofit schemes. System specifications were extracted from detailed as-built retrofit design drawings and subsequently input into the EnergyPlus simulation platform. The primary objective was to quantitatively assess the effectiveness of the proposed EEMs.

As depicted in Figure 6, the simulation results reveal a substantial reduction in energy consumption for both lighting and HVAC systems following the implementation of the retrofit strategies. These findings underscore the energy-saving potential of the proposed interventions and affirm their applicability in enhancing the energy performance of existing small-to-medium-sized office buildings.

For the lighting system, the annual electricity consumption decreased from 16,438 kWh to 8482 kWh, representing a 48% reduction. Regarding HVAC systems, two retrofit schemes were simulated.

The first scheme focused on upgrading the performance of the existing FCU system, leading to a reduction in annual electricity consumption from 72,737 kWh to 43,853 kWh—an energy saving of approximately 40%. The second scheme involved replacing the existing system with a VRF system, which further reduced annual consumption to 37,543 kWh, corresponding to a saving of approximately 48%. As shown in Figure 7, the simulation results reveal distinct performance characteristics under varying load conditions. Under full load, the total system COP of the FCU system reached 2.56, but its annual average total system COP decreased to 1.76, with monthly values ranging from 1.30 to 2.16. In contrast, the VRF system achieved a higher full-load total system COP of 2.87 and a more stable annual average of 1.84, with the monthly total system COP values ranging from 1.53 to 2.09. The most pronounced efficiency gap was observed during winter months, where the VRF system maintained significantly better performance. This is attributed to its multi-unit inverter design, which enables partial load operation by shutting down selected outdoor units, thereby enhancing part-load efficiency. Conversely, the FCU system utilizes a single chiller, which limits its ability to operate efficiently under low-load conditions.

The monthly electricity consumption patterns further support this distinction. The FCU system exhibited an average monthly electricity use of 2895 kWh, peaking at over 4200 kWh in August. The VRF system consumed less energy overall, with an average of 2654 kWh per month and summer peaks remaining below 3900 kWh, alongside significantly lower consumption during winter. In summary, the VRF system demonstrated superior energy performance and operational flexibility across all seasons, particularly under part-load conditions, making it a more effective solution for buildings with variable thermal loads.

When considering the total building energy consumption, which includes both lighting and HVAC systems, the FCU retrofit scheme reduced the total annual electricity usage from 89,174 kWh to 52,334 kWh, achieving overall energy savings of approximately 41%. In contrast, the VRF scheme further decreased the total electricity consumption to 46,025 kWh, resulting in total energy savings of 48%.

4.3. Renewable Energy Generation

To assess the performance of the PV system, the PV module simulation model within EnergyPlus was utilized. Based on the input parameters listed in Table 5 and the TMY weather data for Kaohsiung, Taiwan, the system’s annual electricity generation was estimated. Considering the actual installed area and module efficiency, the simulation results—presented in Figure 8—indicate that the PV system achieved its highest monthly output in June (1334 kWh) and the lowest in January (954 kWh), resulting in a total annual electricity generation of approximately 13,647 kWh.

Although the rooftop PV system contributes to the building’s energy supply, the total annual output is insufficient to meet the demand of either retrofit scheme. The projected energy shortfalls were 38,687 kWh for the FCU scheme and 32,377 kWh for the VRF scheme. To fully compensate for these deficits, an additional 145.1 m² of PV panels would be required for the FCU scheme and 121.5 m² for the VRF scheme. While the rooftop area is technically sufficient to accommodate the additional panels, potential shading from adjacent buildings may reduce system efficiency. Consequently, system expansion was not pursued in this study, following consultation with the building owner.

4.4. nZEB Performance After Retrofit

Figure 9 presents the roadmap toward achieving nZEB performance for both the FCU and VRF retrofit schemes. In this analysis, the annual energy savings from lighting and HVAC upgrades, along with the renewable energy contributions (rooftop PV and green power purchase), were converted into EUI to align with the building energy labeling standards outlined in Table 1.

Prior to the retrofits, the combined annual electricity consumption for the lighting and HVAC systems corresponded to an EUI of 141.3 kWh/m²·yr, classified as Grade F in Taiwan’s energy labeling system. Upgrading the lighting system alone reduced the EUI by 12.6 kWh/m²·yr, although the building rating remained at Grade F. HVAC improvements further reduced the EUI by 45.8 kWh/m²·yr in the FCU scheme (upgraded to Grade D) and 55.8 kWh/m²·yr in the VRF scheme (upgraded to Grade C).

The integration of the rooftop PV system contributed an additional EUI reduction of 21.6 kWh/m²·yr, enhancing the energy performance to Grade B for the FCU scheme and Grade A⁺ for the VRF scheme. Furthermore, the adoption of a green electricity procurement mechanism supplied 28,388 kWh annually, equivalent to a further EUI reduction of 45.0 kWh/m²·yr. When accounting for both on-site renewable energy and purchased green power, both retrofit schemes ultimately achieved the highest efficiency rating of Grade A⁺.

Post-retrofit, the total EUI for the FCU scheme was reduced from 141.3 to 16.3 kWh/m²·yr, while the VRF scheme achieved a reduction to 6.3 kWh/m²·yr. These results highlight the significant potential for energy savings and the feasibility of meeting net-zero targets.

According to Taiwan’s nZEB criteria for existing buildings, achieving an A⁺ energy label and ensuring that energy-saving technologies account for at least 50% of the total energy reduction is mandatory. The simulation results demonstrated that energy-saving contributions were 65% for the FCU scheme and 77% for the VRF scheme, both exceeding the threshold. In conclusion, the proposed retrofit strategies not only substantially improved energy performance but also demonstrated the practicality of integrating renewable energy and green power mechanisms—offering a replicable model for future building energy upgrades and decarbonization efforts.

5. Conclusions

This study proposes a comprehensive assessment framework for retrofitting existing small- to medium-sized office buildings into nZEBs. The framework integrates diagnostics of existing conditions, evaluation of technical benefits, alignment with policy standards, and resource integration strategies. It demonstrates not only strong policy relevance but also practical applicability. The proposed framework serves as a technical support tool for government-led energy retrofit initiatives and assists designers and building operators in establishing a clear decision-making process and technology selection criteria, thereby enhancing retrofit efficiency and effectiveness.

This study used a bank branch office in Kaohsiung as a case study to evaluate the effectiveness of various retrofit strategies through on-site investigations and energy simulations. The results show that implementing a high-efficiency fan coil unit (FCU) system can reduce the energy use intensity (EUI) from 141.3 to 82.9 kWh/m²·yr. With the addition of rooftop photovoltaics and green power procurement, the EUI further decreases to 16.3, achieving an A⁺ energy performance rating and exceeding 50% energy savings to meet Taiwan’s nZEB definition. An alternative scenario using a variable refrigerant flow (VRF) system demonstrated even better performance, with an EUI reduced to 6.3 under the same renewable configuration, approaching net-zero levels. However, higher upfront costs and installation complexity may hinder its practical adoption. It is recommended that the government support mechanisms—such as subsidies or Energy Service Company (ESCO) models—be introduced to lower these barriers. Future research should incorporate long-term monitoring and user behavior analysis to validate simulation results and strengthen the framework’s applicability in real-world conditions.

The green power procurement mechanism is particularly noteworthy. In dense urban environments, small- to medium-sized buildings often face physical limitations that restrict the installation of sufficient on-site renewable energy systems. Green power procurement enables such buildings to overcome spatial constraints by sourcing renewable electricity through the power market, thus offering a viable pathway to meet nZEB goals. Strengthening market mechanisms and policy incentives in this area could further encourage participation from the commercial building sector.

It is important to acknowledge that the findings of this study are based on simulation results and currently lack post-retrofit operational data. Since the retrofit project has only recently been completed, at least one full year of operation is required before sufficient empirical data can be collected. In the future, we will integrate the deployed energy monitoring system in this case study, which provides historical energy consumption data of major HVAC equipment at five-minute intervals. On-site measurements will also be conducted to verify the accuracy of the monitoring data. This approach will enable the long-term tracking and empirical evaluation of the energy performance of HVAC equipment and the electricity generation efficiency of the photovoltaic system. Furthermore, a comprehensive comparison will be made between the electricity bills for a full year after the retrofit and those from the corresponding period before the retrofit to validate the applicability and effectiveness of the proposed evaluation framework. The post-retrofit validation process will generally follow the procedures and methods established during the retrofit phase.

In conclusion, this study verifies the feasibility and scalability of the proposed nZEB retrofit framework for small- to medium-sized existing office buildings in Taiwan. It provides a practical and actionable roadmap for deep energy retrofits and offers valuable insights to support Taiwan’s transition toward low-carbon buildings and the achievement of the 2050 net-zero target.