Szu wei Fang

Dr. Szu wei Fang

Infrastructure Planning and Engineering, Feng Chia University, Taichung, Taiwan
orcid

Research Focus & Metrics

Academic Career Length

2Years (2025-Present)

(Calculated once Scilit publication data is synced: current year − year of first publication + 1)

Representative Publications

View All Publications
18 August 2026 in Sustainability
Verified
Journal Article
Open access
PDF file
Po-Chuan HuangPo-Chuan Huang
Po-Chuan Huang
SFSzu wei Fang
Szu wei Fang
WLWei Lin
Wei Lin
SLShwu-Ting Lee
Shwu-Ting Lee
Existing residential buildings require peak-demand governance even when complete sub-metering and advanced control systems are unavailable. This study develops a replicable asset-light demand-side management (DSM) protocol for contracted-capacity risk assessment. The protocol integrates utility-billing records, 15 min community main-meter (MP) and selected sub-meter data, operational-period segmentation, residual-load treatment, capacity-risk indicators, rule selection, event decomposition, and post-calibration verification. The protocol was examined in a Taiwanese condominium using utility records beginning in June 2022 and stabilized digital-meter data from April 2023 to May 2026. Four sub-metered systems explained 76.3% of monthly MP electricity use, with 23.7% retained as residual load. Under a 40 kW operational reference boundary, maximum MP demand reached 41.12 kW, yielding a utilization ratio of 1.028 and a relative capacity buffer of −2.8%. In a representative critical interval, the aggregate mechanical-power boundary contributed 79.6% of MP demand; however, equipment-level EV output was not independently recorded. After rule-based calibration, maximum MP demand was 29.36 kW, and no MP threshold-exceedance intervals or billing-level over-capacity surcharges were documented. These findings provide observational performance evidence rather than definitive causal proof. The contribution is a recalibratable diagnostic sequence for capacity-risk governance under partial metering and property-management constraints.
1

19 November 2026 in Sustainability
Verified
Journal Article
Open access
PDF file
HLHAN CHUN LIU
HAN CHUN LIU
SFSzu wei Fang
Szu wei Fang
WLWei Lin
Wei Lin
SLShwu-Ting Lee
Shwu-Ting Lee
Po-Chuan HuangPo-Chuan Huang
Po-Chuan Huang
Solar-Term Architecture treats time as an active design variable; this article develops its measurable interpretation through the In-SENSE Lab, an adaptively reused office in Taichung, Taiwan. Continuous thermal and CO2 monitoring is combined with companion-study spot measurements of lighting and noise, indexed by the Twenty-Four Solar Terms as a culturally grounded calendar for locally measured environmental change rather than as fixed climatic labels. Term-level analysis uses 6561 laboratory hourly means across 19 Solar Terms (1 December 2025–4 September 2026). The companion study reports an annual mean predicted mean vote of −0.16, mean weekday CO2 of 465 ppm, a color-rendering index of 88–98, background noise of 26.2 dB(A), and eight of twelve months without mechanical cooling. Term statistics resolve transitions hidden by monthly aggregation: occupied-hour CO2 medians reach 565–785 ppm in adequately sampled terms; uncorrected fixed-input PMV medians move from negative in winter to near zero in summer (winter values extend beyond the ±3 scale of ISO 7730 and are read as within-case indicators); and metered daily energy separates free-running days (about 1 kWh/day) from cooling episodes (up to 117.0 kWh/day). Annual averages alone are insufficient. The principal contribution is a replicable four-domain protocol linking Solar-Term temporality with indoor environmental quality, and a minimum term-level reporting template.


24 December 2025 in Buildings
Verified
Journal Article
Open access
PDF file
WLWei Lin
Wei Lin
SFSzu wei Fang
Szu wei Fang
SLShwu-Ting Lee
Shwu-Ting Lee
This study proposes a scalable cyber–physical system (CPS) framework utilizing a hierarchical five-layer architecture to enhance indoor environmental quality and energy efficiency. The methodology integrates a Random Forest-based predictive model trained on a 22-month longitudinal dataset (2024–2025) to separate climatic effects from occupancy-driven loads. This study prioritized the development of a high-precision and cost-effective monitoring architecture to address the persistent challenge of sustaining thermal comfort in subtropical academic laboratories. The proposed system achieved a validation mean absolute percentage error (MAPE) of 2.50%, indicating strong predictive reliability. Hardware expenditures were below USD 400, substantially reducing barriers to broader adoption. Field deployment confirmed an operational EUI of 188.6 kWh/m2·year, which is 28.5% lower than prevailing regional benchmarks, while consistently meeting stringent indoor air quality (IAQ) requirements. Additionally, simulation modules calibrated with the validated dataset indicated a further 15–20% reduction potential through the application of active control strategies. Collectively, these findings establish a transferable empirical reference for climate-responsive operational practice.