Optimizing Electric Bus Efficiency: Evaluating Seasonal Performance in a Southern USA Transit System

Abstract

Electric buses (EBs) are increasingly adopted for their environmental and operational benefits, yet their real-world efficiency is influenced by climate, route characteristics, and auxiliary energy demands. While most existing research identifies winter as the most energy-intensive season due to cabin heating and reduced battery performance, this study presents a contrasting perspective based on a three-year longitudinal analysis of the LYMMO fleet in Orlando, Florida—a subtropical U.S. region. The findings reveal that summer is the most energy-intensive season, primarily due to sustained HVAC usage driven by high ambient temperatures—a seasonal pattern rarely reported in the current literature and a key regional contribution. Additionally, idling time exceeds driving time across all seasons, with HVAC usage during idling emerging as the dominant contributor to total energy consumption. To mitigate these inefficiencies, a proxy-based HVAC energy estimation method and an optimization model were developed, incorporating ambient temperature and peak passenger load. This approach achieved up to 24% energy savings without compromising thermal comfort. Results validated through non-parametric statistical testing support operational strategies such as idling reduction, HVAC control, and seasonally adaptive scheduling, offering practical pathways to improve EB efficiency in warm-weather transit systems.

1. Introduction

As cities transition to sustainable transportation, electric buses are becoming integral to reducing emissions and improving energy efficiency. However, optimizing their performance remains challenging due to various operational and environmental factors. Central Florida’s transit system faces various challenges due to its mix of urban and suburban areas, with considerable distances between destinations, imposing significant stress on bus fleets and leading to high mileage. This causes higher fuel usage and is responsible for higher costs and emissions. Multiple alternative fuels are being explored to address these challenges around the US, with battery Electric Buses (EBs) emerging as a popular option [1]. In downtown Orlando, a highly congested area, battery-powered buses are deployed to optimize mobility, energy efficiency, and performance. However, several obstacles [2], such as traffic conditions, high tourist traffic, range limitations, charging time, maintenance requirements, driving cycles, and uncertainties associated with this bus type, may hinder the widespread adoption of EBs. Additionally, installing expensive electric vehicle charging stations and charging infrastructure upgrades pose additional challenges, especially with an expanded fleet’s increased load and demand [3].

Various public transit stakeholders in Central Florida and neighboring areas have begun exploring and have already introduced small-sized EB fleets. Examples include [4] BEEP’s battery-powered autonomous shuttles in Altamonte Springs and Lake Nona, Central Florida Regional Transit Authority’s (LYNX) EB fleet in downtown Orlando [5], and Orange County’s electric school buses [6]. In South Florida, approximately 200 miles south of Central Florida, Broward County is set to deploy 60 EBs [7] in the 2023/2024 school year, facilitated by the state’s Department of Environmental Protection. Many other regions and municipalities are considering implementing EBs in their public transportation systems.

Central Florida features a subtropical to tropical climate characterized by distinct seasonal variations [8]. The state encounters hot and humid summers, where temperatures frequently surpass 90 °F (32 °C) and humidity remains high. Afternoon thunderstorms are common during this period owing to Florida’s proximity to the Gulf of America and the Atlantic Ocean. Winters, particularly in the southern regions, offer mild weather with temperatures typically ranging between the 60s and 70s °F (15–25 °C). Much of the state is characterized by low-lying terrain and renowned flatness, with its highest natural point at Britton Hill reaching merely about 345 feet (105 m) above sea level [9].

Electric buses exhibit varying energy efficiency, influenced predominantly by several factors [10] encompassing vehicle-related parameters such as the technical specifications of the bus, the battery characteristics, and the Heating, Ventilation, and Air Conditioning (HVAC) system. Operational parameters, such as driving style, passenger loads, initial state of charge (SOC), and braking style, also significantly impact energy efficiency. Route-related parameters, specifically road gradients, contribute to energy consumption. Moreover, environmental factors like ambient temperature and traffic conditions play a role [11]. Considering all factors, the energy consumption of a typical EB tends to be higher in winter due to the increased usage of heating to maintain comfortable interior temperatures [12,13].

Although several studies have examined EB energy consumption under varying environmental conditions, most conclude that winter poses the most significant energy challenge due to battery inefficiency and heating demands. However, these findings are often based on cold or temperate climates and lack region-specific analysis in warmer areas. There remains a clear gap in understanding how subtropical conditions, such as those in Florida, influence EB performance, particularly the combined effects of high ambient temperatures, increased HVAC loads, and congestion.

The primary objectives of this study include investigating the influence of multiple factors on the energy efficiency of EBs. The factors include speed, idling, driving patterns, and HVAC usage, among others. For this study, the research team partnered with the LYNX [14], particularly its EB fleet division known as “LYMMO” [15], which operates within Downtown Orlando. LYMMO is a complimentary downtown circulator that offers free transportation services to its users. The LYNX LYMMO service was developed in 1997 to provide free access to public transportation in the downtown business, entertainment, and shopping district. The three LYMMO Downtown Lines (Orange, Grapefruit, and Lime) have dedicated lanes and control their traffic signals along routes to optimize performance. The LYMMO Orange Line provides frequent service to specific locations in downtown Orlando, operating every 8 min during office hours and every 15 min during evenings, weekends, and holidays. The LYMMO Grapefruit Line operates every 8–10 min during office hours and every 15 min during evenings, weekends, and holidays. The LYMMO Lime Line operates every 15 min during office hours and every 20 min during evenings, weekends, and holidays. All the routes and operating hours mentioned here were last updated in May 2024. LYMMO buses began transitioning to an all-electric fleet in October 2020 to be fully electric by the end of Summer 2023 with 14 EBs.

Central Florida presents a unique case for the evaluation of electric buses due to its subtropical climate, characterized by high temperatures and humidity, which can significantly influence energy consumption, particularly through the increased demand on HVAC systems. Moreover, the region’s rapid population growth, fueled by tourism, results in high traffic congestion in urban and downtown areas. This adds another layer of complexity to electric bus operations, making it an ideal setting to explore how these factors—climatic and traffic-related—interact to impact bus performance. The region’s conditions make it especially relevant for studying the efficiency of electric buses in an environment that is both demanding and representative of other subtropical regions.

This research investigates the feasibility of EBs in Central Florida, where their performance has yet to be established. Transit authorities need to understand the fleet’s outcomes before expanding. The research focuses on a comprehensive energy efficiency evaluation and long-term performance evaluation, considering seasonal variations and annual trends. The research team gathered detailed data during two distinct periods: June 2022 (summer) and January 2023 (winter), and the analysis encompasses a one-year timeframe from June 2022 to June 2023.

2. Literature Review

Within the ongoing discussions about the environmental impact and the infusion of renewable energy into public transportation, the critical focal points revolve around embracing sustainable practices and maximizing energy efficiency. To minimize greenhouse gas emissions and enhance energy utilization efficiency, transit authorities are exploring alternative fuels to replace traditional diesel in buses [16]. EBs, encompassing battery-powered and plug-in hybrid variants, have emerged as compelling and economically feasible options for fostering sustainable transit solutions. This strategic shift aligns with the broader goal of enhancing environmental stewardship and operational efficiency in the public transportation sector [17].

In Florida, the Department of Transportation (FDOT) emphasizes the need to prioritize rural, underserved, and disadvantaged communities for electric vehicle (EV) charging infrastructure [18]. However, despite the increasing interest in EBs, several obstacles hinder the global electrification of public transportation. These include range anxiety, lengthy charging times, higher upfront procurement costs, and operational challenges [19].

Various states are actively investigating potential implementations of EBs in the United States. Notably, California has conducted a feasibility study based on daily operations, charging infrastructure, energy consumption, and life cycle costs, as it leads the nation in EV sales, accounting for over half of all EVs sold in the country [20]. One crucial aspect of EBs is their long-term performance and durability analysis [21]. Energy consumption [22] plays a vital role in understanding overall efficiency. Optimizing parameters like driving habits, idle periods, weather circumstances, and accurate passenger traffic forecasts led to a notable reduction in energy consumption [23].

Numerous research endeavors unravel the significant factors influencing the energy consumption of EBs [24]. These factors can be roughly categorized [25] into vehicle-related [26] attributes, operational considerations, and environmental factors. The first encompasses technical specifications, battery intricacies, HVAC, and auxiliary systems. Operational considerations [27] include the details of driving styles, passenger loading dynamics, initial state of charge, and energy recuperation through braking. The impact of the route has been explored [11,28], focusing on stop spacing and the consequential influence of road gradient. Environmental factors [29], especially ambient temperature, are also important as they substantially affect the usage of HVAC systems [30]. Driving behavior significantly impacts energy consumption and range utilization [31]. Studies often adopt classification methods to identify driving styles and compare energy consumption characteristics under different driving conditions [32]. Engaging in aggressive driving, characterized by frequent acceleration and deceleration, leads to increased power loadings and energy consumption compared to moderate or eco-driving, characterized by smooth acceleration and deceleration while maintaining a relatively constant speed [33,34]. This distinction is more prominent during short-distance urban trips [35]. Optimizing driving behavior, such as maintaining a constant speed between stops and minimizing acceleration and deceleration, can extend the driving range and maximize energy efficiency. Minimizing unnecessary idling is critical to optimize the efficiency and range of EBs. Energy-efficient heating and cooling systems, intelligent energy management to shut down non-essential systems during idling, and encouraging drivers to turn off the bus when parked can reduce energy consumption during operation and improve overall energy efficiency [36].

HVAC systems represent one of the most significant contributors to auxiliary energy consumption in electric buses, substantially affecting operational efficiency and vehicle range. Fiori et al. showed that HVAC usage can reduce the electric vehicle’s driving range by up to 24% during heating and 10% during cooling operations, highlighting its critical role in energy management [37]. To address inefficiencies in conventional HVAC monitoring, Guo proposed a real-time energy performance benchmarking method that uses autoencoder neural networks and Gaussian Process Regression to identify underperforming AC systems across large urban fleets, enabling adaptive, data-driven fault detection and optimization [38]. This adaptive framework supports continual learning, making it suitable for integrating smart city platforms and growing EV fleets.

The interaction between HVAC systems and idling patterns is particularly relevant in transit bus operations. A report by BusRide notes that HVAC systems are the largest identifiable source of battery drain during idle periods, with additional subsystems (e.g., surveillance, lighting, and fare collection) contributing up to 10–20% of idle energy usage [39]. Jia further expands on HVAC’s role in overall energy management by developing a deep reinforcement learning-based control strategy that integrates thermal comfort and road-grade anticipation, achieving energy savings while maintaining cabin comfort [40]. In a follow-up study, Jia introduced a predictive framework incorporating passenger load forecasting into real-time HVAC and powertrain control, demonstrating improved energy efficiency in dynamic urban environments [41].

Recent works by Jahic and Deptula emphasize the importance of multi-factor and AI-driven approaches in broader modeling contexts. Jahic comprehensively reviews energy consumption factors, including environmental and operational variables. Deptula introduces expert systems and multi-valued logic trees to evaluate electric vehicle energy use under varying conditions [42,43]. These methods collectively support a shift toward holistic and intelligent EMS designs. HVAC systems are optimized with propulsion and auxiliary subsystems to enhance range, passenger comfort, and fleet-wide efficiency.

Many studies have investigated the energy usage of battery electric buses in different regions, climates, and operating conditions. In California, 40 ft electric buses were reported to consume between 2.4 and 2.8 kWh/mile, depending on the season, while 60 ft models ranged from 3.3 to 4.1 kWh/mile [44]. In European cities, BEBs showed a wider range of 1.24 to 2.48 kWh/km, highlighting differences in city-specific traffic and route profiles [45]. In New Zealand, seasonal variation was also evident, with buses consuming approximately 1.2–1.4 kWh/km depending on ambient conditions and auxiliary loads [46]. A Canadian study found that the operational fleet demand increased significantly in winter, with a replacement factor rising to 1.52 compared to 1.21 in summer [47]. In California again, Proterra buses consumed 1.34 kWh/km on average [48]. Another European-wide analysis indicated usage between 14 and 16 kWh/h, translating to around 0.84–0.96 kWh/km based on fleet speed [20]. Finnish data further supports seasonal sensitivity, with consumption increasing from 1.24 to 1.30 kWh/km in summer to 1.71–1.95 kWh/km in winter [12]. Notably, a study in Latvia demonstrated that energy consumption could be reduced to as low as 0.42 kWh/km by implementing semi-dynamic charging and energy recuperation strategies [49]. In the same research, it was mentioned that German researchers reported seasonal variation from ~2.1 kWh/km in summer to 4.1 kWh/km in winter, with 1.7 kWh/km attributed to HVAC and comfort systems. Similarly, Spanish findings indicated load-based variation, with energy use rising from 3.61 kWh/km under low loads to 4.59 kWh/km under high loads [50].

Table 1. Summary of energy consumption findings for electric buses in various studies.

Location	Bus Type	Seasonal Calculation	Findings
California, USA	BEB	Seasonal (All Seasons)	40 ft: 2.6 ± 0.4 kWh/mile; 60 ft: 3.6 ± 0.5 kWh/mile; seasonal variation observed
Germany and Denmark	BEB-BYD	Not Mentioned	1.24 to 2.48 kWh/km in European cities
Auckland, New Zealand	N/M	Seasonal (Winter vs. Non-Winter)	Winter: 43 kWh/h; Non-Winter: 36 kWh/h (est. 1.2–1.4 kWh/km)
Toronto, Canada	BYD, New Flyer, Proterra	Seasonal (All Seasons)	Replacement factor ranges from 1.2 to 1.52 across seasons
California, USA	Proterra	Not Mentioned	Average energy consumption: 1.34 kWh/km
Europe	N/M	Not Mentioned	Increased from 14 to 16 kWh/h during colder temperatures (Speed: 16.7–19.1 km/h)
Finland, Europe	BEB	Seasonal (Winter Only)	Summer: 1.24–1.30 kWh/km; Winter: 1.71–1.95 kWh/km
Latvia	N/M	Not Mentioned	Reduced to 0.42–0.99 kWh/km with semi-dynamic charging
Germany	N/M	Seasonal (Summer vs. Winter)	Summer: 2.1 kWh/km; Winter: 4.1 kWh/km (HVAC use: 1.7 kWh/km)
Spain	N/M	Not Mentioned	3.61 kWh/km (low load) to 4.59 kWh/km (high load)
Rio de Janeiro, Brazil	BEB	Not Mentioned	1.69 kWh/km
Kuala Lumpur, Malaysia	BEB	Not Mentioned	5.36 kWh/km

provides a summarized overview of energy consumption findings for electric buses reported across different global studies. A study in Rio de Janeiro, Brazil, reported an electric bus energy consumption of 6.1 MJ/km, equivalent to approximately 1.69 kWh/km [51]. Similarly, a study in Kuala Lumpur, Malaysia, found that in the most efficient scenario (Scenario R), electric buses consumed 5.36 kWh/km, outperforming the other four operational scenarios [52].

Understanding the impact of environmental conditions on battery capacity is crucial. High temperatures can affect the fast charger’s functionality and reduce the battery’s lifespan, while low temperatures can decrease battery power and capacity, affecting the vehicle’s range and overall cost [53]. Higher wind speeds and lower temperatures are associated with increased energy consumption and a poor energy regeneration rate [54]. In this context, accurately measuring and estimating electricity usage becomes essential for improving electric buses’ overall energy efficiency. A core technological advantage of EVs over internal combustion engine (ICE) vehicles is integrating regenerative braking systems (RBS), which capture and convert kinetic energy into electrical energy during deceleration [55]. This is accomplished by reversing the electric motor to generate negative torque at the drive wheels, thereby recharging the battery [56].

This energy recovery mechanism is particularly beneficial in urban environments, where frequent stop-and-go conditions allow RBS to improve energy efficiency significantly. As a result, EBs tend to perform more efficiently in interrupted urban driving cycles than uninterrupted highway travel. In contrast, ICE vehicles experience reduced fuel efficiency in urban conditions due to frequent braking and the inherent thermal losses of combustion engines [57]. Coupled with environmental factors, these innovations shape the operational efficiency and real-world viability of alternative-fueled transit systems.

While extensive research on energy analysis and performance evaluation has examined the impact of ambient temperature on EB energy consumption, most studies have focused on cold climates, where heating demands increase energy use. However, fewer studies have explored EB performance in hot and humid regions, specifically tropical and subtropical regions, where continuous HVAC operation can significantly impact efficiency. A gap remains in studying the Southern region, particularly Florida, where high congestion and relatively mild yet persistent weather conditions may influence EB performance. Conducting specialized investigations tailored to this specific geographical and weather context is essential to comprehend how these factors influence energy consumption and overall performance. By undertaking such research, valuable insights can be gained to advance EB technology in Central Florida and strategically integrate EBs into the existing transit infrastructure.

In conclusion, the popularity of EBs is on the rise [58], and they are increasingly considered a viable alternative to conventional buses for reducing emissions and improving air quality [59]. However, addressing the research gaps in the context of Central Florida’s specific congestion pattern and geographical and weather conditions is crucial to optimizing energy efficiency and performance and successfully implementing EBs in the region’s public transportation system.

3. Methodology

The data for this research is derived from the LYNX LYMMO fleets, and the primary tool employed for data collection is the ViriCiti Electronic Logging Device (ELD) [60]. It is commonly known as a telematics device. The telematics device plays a central role in gathering essential data [61,62]. The device uses sensors and GPS technology to track crucial parameters such as location, speed, battery information, energy consumption in different states, codes, messages, and the driving behavior of vehicles [63]. Telematics devices are equipped with reliable hardware capable of collecting high-frequency GPS fleet-tracking data, ensuring a comprehensive and granular dataset for analysis. The ViriCiti ELD continuously transmits information about the buses to the central server, and the ChargePoint platform monitors them. Figure 1 shows a LYMMO bus unit at LYNX Central Station in Orlando, Florida, and Figure 2 depicts the route map of LYMMO operated in downtown Orlando by LYNX.

A combination of real-time and historical data from the EBs’ on-board devices was utilized to assess the energy efficiency and performance of the buses. The ELD is permanently logged with the bus shown in Figure 3. The data was accessed through an online dashboard called ViriCiti, shown in Figure 4. ViriCiti offers data and visualization tools that aid in the analysis process. For this study, the focus was on several vehicle parameters, including (1) the rate of energy consumption (kWh/km) and (2) total energy consumption (kWh). The energy consumption data includes both the energy consumed during driving and the energy consumed during idling, (3) distance traveled (km), (4) trip duration (hours), and (5) average speed (km/h). The hourly average speed was considered for analysis in this study, as well as (6) HVAC temperature inside the bus, (7) ambient temperature, and (8) the state of charge.

EBs efficiency can be interpreted from the fuel economy. Fuel economy is measured in kWh/mile, where an EB with a fuel efficiency of 2.17 kWh/mile equals 17.35 miles per Diesel Gallon Equivalent (DGE) [65]. EBs exhibit higher fuel efficiency than diesel buses due to the incorporation of regenerative braking [66]. Out of many battery-related information, the SOC is closely monitored. The SOC denotes the remaining energy in a battery pack, analogous to a fuel gauge. Although theoretically ranging from 0% to 100%, practical SOC limits are set to safeguard long-term battery life (48). The lithium-ion-based batteries are widely used in EBs for power and higher density [67]. In general, EB batteries exhibit a wide capacity range, varying from 76 kWh (e.g., Nova bus LFSe) to 660 kWh (e.g., Proterra Catalyst E2) (51). The Proterra fleet of LYNX LYMMO in Orlando, Florida, opts for the Lithium Iron Phosphate battery with a 425 kWh energy capacity, providing a maximum range of around 220 miles. A full off-route recharge takes up to five hours. The dataset spans from June 2021 to December 2024, capturing over three years of high-resolution operational data from a fleet of 14 electric buses. Notably, six of these buses were added in 2023, which results in staggered data availability across the study period. This dataset was collected to support a detailed investigation into real-world electric bus performance, with specific attention to energy efficiency, seasonal variation, regenerative braking effectiveness, and auxiliary energy demand related to heating, ventilation, and air conditioning systems.

Each entry in the dataset represents a one-hour operational snapshot of a particular bus. The recorded parameters include average driving speed, total energy consumed, energy regenerated during braking, distance traveled, SOC metrics, and time spent in various operational states such as driving, idling, charging, and in service. These values are aggregated from raw telemetry into hourly average or cumulative summaries, making the data both fine-grained and analytically manageable.

The preprocessing phase began with consolidating monthly data exports from the ViriCiti platform. The raw data, initially stored in a wide format with each metric–bus pair as a separate column (e.g., “297-302—Energy used”), was reshaped into a long format. In this format, each row contained a single metric for a given time and bus ID, which enabled efficient filtering and aggregation. Following validation and standardization, the dataset was divided into a wide format, producing one row per hour per bus, with separate columns for each operational parameter.

To enhance the dataset’s usability, the combined timestamp was split into two distinct columns: date and time (hours), with the latter expressed as a decimal-based hour (e.g., 13.5 representing 1:30 p.m.). A new season variable was also added based on standard meteorological classifications: winter (Dec–Feb), spring (Mar–May), summer (Jun–Aug), and fall (Sep–Nov). This structure supports both time-of-day and seasonal analyses across the entire fleet.

Special care was taken in the treatment of missing values. Gaps during early-morning hours (e.g., 1–3 a.m.) were preserved as indicators of non-operational time, not treated as data loss. Similarly, their absence in earlier records was acknowledged as valid for buses introduced in mid-2023 (such as IDs in the 297–320 and 384–389 ranges). These entries were left untouched, preserving the true operational history of each vehicle.

An additional dataset containing HVAC energy usage proxies was later integrated. This dataset, preprocessed and time-aligned, was merged into the fleet dataset using a composite key of date, time (hours), and BusID. As a result, each record now includes both propulsion-related and auxiliary energy data, allowing for a holistic analysis of hourly energy demands.

The final cleaned dataset contains over 135,000 hourly records covering 14 buses and more than 30 operational parameters. It enables a wide range of analytical tasks, including time-of-day and seasonal performance profiling, energy consumption benchmarking, regenerative braking analysis, and statistical hypothesis testing using non-parametric methods such as the Kruskal–Wallis H-test. The structured and well-documented methodology ensures that all findings are reproducible, scientifically rigorous, and representative of real-world electric bus operations.

4. Statistical Analysis and Results

4.1. Monthly Trends in Transit Fleet Energy and Operational Metrics (2021–2024)

To uncover long-term temporal trends and shifts in operational behavior, the dataset was aggregated to monthly averages across the years 2021 to 2024. Eight key metrics were analyzed, covering energy usage, regenerative braking efficiency, vehicle speed, and time-in-service indicators. The resulting time series plots (Figure 5) reveal several noteworthy patterns:

Both energy used and SOC used exhibit clear seasonal cycles, with notable peaks in the summer months (June–August) and secondary elevations in early winter. These trends likely reflect increased HVAC demand during temperature extremes and potential increases in passenger volume during holiday periods. The upward trend in energy use across years suggests increasing service demand, system inefficiency, or the growing deployment of electric buses across more routes.

The regenerative braking metric shows more stable trends, with mild increases during summer months and dips in the winter. This may be due to smoother braking behavior in cooler traffic conditions and longer service routes.

Distance driven reflects the monthly service volume. Notable dips are visible in early 2022 and again during late 2023. A pronounced drop is observed in 2024, attributed to operational limitations: several electric buses experienced service issues, prompting the transit authority to restrict the deployment of EBs within the fleet temporarily. This action directly impacted the total distance and energy utilization during the affected months.

The average speed metric remains relatively stable but displays occasional seasonal drops, especially during mid-year and winter breaks. These drops may correspond to increased congestion or shorter intra-campus shuttle loops during peak semester activity. This aggregated overall consumption metric reflects a compound view of fleet load and follows trends similar to energy usage and distance. Variations across months and years reaffirm the operational and environmental influence on transit system demand.

The time idling and time in service metrics consistently range between ~0.5 and 0.85 h per operational hour, indicating that, on average, a bus spends ~30 min idling and ~50 min in active service per recorded hour. These values align with typical transit operational behavior, where downtime, partial hour runs, or traffic-related delays affect total time utilization.

4.2. Seasonal Variation Analysis

To understand temporal and seasonal trends in transit bus performance, the smoothed hourly averages of three key metrics—energy used, energy regenerated during driving, and average speed—were plotted by season. The results are presented in Figure 6.

Across all seasons, energy consumption peaks between 12 p.m. and 5 p.m., coinciding with operational demand and possibly higher passenger load or HVAC usage during warmer hours. Summer exhibits consistently higher energy usage compared to other seasons, likely due to increased cooling demand and potentially higher service frequency. Winter shows relatively lower energy usage, suggesting reduced vehicle demand or milder operational loads.

Regenerative energy recovery exhibits an inverse pattern to energy use, with notable dips during midday hours and peaks during the early morning (6–9 a.m.) and late evening (8–10 p.m.). Fall and spring show the most effective regeneration patterns, possibly due to more stable and moderate driving conditions. Summer regeneration is relatively suppressed, suggesting more aggressive driving or braking inefficiencies, or HVAC overhead reducing regeneration efficiency.

The average speed remains fairly stable across seasons, ranging between 18 and 25 km/h, but slight variations exist. Spring and fall support marginally higher speeds, while winter and summer experience more fluctuations, possibly due to operational constraints like weather or traffic patterns. Speed generally drops during midday, suggesting congestion during peak service hours.

To further explore the distribution of energy and performance-related variables across seasons, Gaussian Kernel Density Estimation (KDE) plots were generated for energy used, energy regenerated during driving, and average speed. These plots reveal non-normal distribution patterns and highlight seasonal shifts in energy intensity and driving behavior.

4.3. Seasonal Distribution Analysis for Energy and Speed

To investigate how key operational metrics vary in distribution across different seasons, Gaussian Kernel Density Estimation (KDE) plots were generated for three selected variables: energy used, energy regenerated during driving, and average speed. These smoothed density plots offer a more nuanced view of the data, highlighting central tendencies, spread, and seasonal shifts as shown in Figure 7.

The Gaussian KDE plots highlight clear seasonal variations across key transit metrics. For energy used, the distribution shifts notably to the right in summer, reflecting a higher energy demand likely driven by increased HVAC usage and operational intensity. In contrast, winter and fall exhibit lower median values, with winter showing the narrowest spread, indicating more consistent but reduced energy consumption. Spring remains intermediate, suggesting moderate operational demand. Energy regenerated during driving shows more symmetrical distributions, though fall and spring present slightly higher densities in the mid-range, implying optimal braking conditions or favorable driving patterns. In summer, regeneration shifts subtly lower, possibly due to longer, smoother driving cycles or HVAC-related battery load. Finally, average speed remains relatively consistent across seasons, yet summer and fall show peaks around 20–22 km/h. Winter and spring exhibit slightly lower speeds, reflecting the effects of cold-weather driving behavior and potential congestion during high-demand periods.

4.4. Seasonal Distribution of Operational Matrix

Boxplots were created to compare distributional differences across seasons for six critical variables: energy used, energy regenerated during driving, average speed, distance driven, time driving, and time idling as shown in Figure 8. These visualizations provide insight into the variability and central tendencies of the fleet’s performance under different environmental and operational conditions.

The energy used shows the highest median and upper quartile in summer, aligning with elevated HVAC loads and possibly higher ridership. Winter demonstrates the lowest spread and median, suggesting reduced energy demand and tighter operational control. Spring and fall are intermediate, with moderate spread and consistent usage profiles.

The energy regenerated during driving reveals relatively balanced distributions across seasons, though spring and winter show slightly higher medians and narrower spreads, indicating favorable regenerative driving conditions. Summer and fall again show a lower range, supporting earlier findings that higher HVAC demand may reduce regenerative effectiveness.

Average speed and distance driven maintains a tight interquartile range, reflecting a stable fleet control across all seasons.

The time driving follows a similar trend to distance, with longer driving times in winter and fall. At the same time, Time idling is noticeably higher in fall, consistent with cold-weather operations requiring longer warm-up times, longer layovers, or traffic-induced delays.

4.5. Seasonal Driving vs. Idling Dynamics

Understanding how electric buses allocate operational time between driving and idling is critical for identifying inefficiencies and optimizing overall fleet performance. Figure 9 presents a stacked bar chart comparing the average time spent in each mode across all four seasons.

Across the board, idling time consistently accounts for most service hours. However, Winter exhibits the highest relative share of idling, which may be attributed to longer vehicle warm-up durations, cold-weather traffic delays, or increased layover times. The rest of the seasons are balanced.

To further explore temporal and seasonal variations, hourly time series plots were generated for energy consumed during driving, energy idled, SOC used driving, SOC used idling, time driving, and time idling, as shown in Figure 10. These plots reveal clear behavioral cycles and seasonal effects. Driving-related energy consumption peaks sharply during midday hours, particularly in summer and spring, aligning with operational load and elevated HVAC usage. The energy idled remains relatively stable but increases slightly during midday and late evenings in summer, likely reflecting extended idling due to cooling demands. SOC usage during driving peaks in summer around midday, while SOC used during idling is low in winter, which means less consumption of HVAC demands.

Temporal patterns in the time spent driving reveal morning and afternoon peaks (8 a.m.–5 p.m.), consistent with commuter periods. Conversely, idling time is more evenly spread throughout the day, with a slight elevation in fall and lowest in spring.

4.6. Statistical Significance of Seasonal Variations

Non-parametric statistical tests were used to rigorously evaluate whether key operational metrics of the electric bus fleet varied significantly by season. The Kruskal–Wallis H-test, a robust method for comparing more than two independent groups, indicated statistically significant differences across all eight key variables (p < 0.05 for each), strongly suggesting that seasonality has a measurable effect on both energy usage and service behavior, as shown in

Table 2. The Kruskal–Wallis H-test results.

Variable	H-Statistic	p-Value	Significant
Energy used	2273.2	0.00 × 100	Yes
Energy regenerated driving	1683.87	0.00 × 100	Yes
Average speed	357.68	3.24 × 10−77	Yes
Consumption overall	7613.77	0.00 × 100	Yes
Distance driven	289.19	2.18 × 10−62	Yes
SOC used	2560.58	0.00 × 100	Yes
Time idling	658.32	2.29 × 10−142	Yes
Time in service	424.81	9.34 × 10−92	Yes

.

Both the energy used and consumption overall exhibited exceptionally high H-statistics (2273.2 and 7613.8, respectively), indicating that energy demand patterns shift markedly between winter and summer. This is consistent with seasonal HVAC loads and fluctuations in passenger demand. The energy regenerated during driving also showed significant variation (H = 1683.87), reflecting seasonal differences in braking behavior and traffic dynamics.

While p-values indicate whether seasonal differences are statistically significant, they do not reflect the magnitude of these differences. To assess the practical significance of our findings, we computed effect size measures for both the Kruskal–Wallis and Mann–Whitney U tests. The effect size provides an interpretable measure of how strongly seasons influence energy consumption, allowing us to evaluate whether statistically significant differences are also meaningful in real-world terms.

Epsilon-squared (ε²) determines a non-parametric effect size for the Kruskal–Wallis H-test. This metric represents the proportion of variance in the dependent variable explained by the independent variable (i.e., season). It is calculated using the following formula:

$${\epsilon }^2={\displaystyle \frac{H-k+1}{n-k}}$$

where

• $H$ is the Kruskal–Wallis test statistic;
• $k$ is the number of groups (here, seasons = 4);
• $n$ is the total sample size.

For HVAC proxy energy consumption, a Kruskal–Wallis statistic of 95.12 was obtained with 4026 observations. Substituting into the formula:

$${\epsilon }^2={\displaystyle \frac{95.12-4+1}{4026-4}}={\displaystyle \frac{92.12}{4022}}\approx 0.229$$

this value (ε² = 0.229) indicates a large effect size, suggesting that seasonality explains a substantial portion of the variation in HVAC energy usage.

The average speed and distance driven were both significantly influenced by the season, with speeds peaking in spring and fall, and the distance increasing slightly during warmer months—suggesting smoother road conditions and potentially expanded service coverage. Time idling and time in service also demonstrated strong seasonal effects; idling time, in particular, was notably higher in winter, possibly due to prolonged vehicle warm-up and increased delays. Finally, the SOC used varied across seasons, with greater battery depletion in summer, driven by heavier HVAC demands and extended duty cycles.

The Bonferroni-adjusted Mann–Whitney U tests provide further granularity by identifying which specific seasonal pairs significantly differ for each operational metric. These pairwise comparisons reveal the depth and directionality of seasonal effects as shown in

Table 3. The Bonferroni-adjusted Mann–Whitney U test outputs.

No.	Metric	Season Comparison	Adjusted p-Value
1	Average speed	Winter vs. Summer	3.09 × 10−18
2	Average speed	Summer vs. Fall	1.61 × 10−8
3	Average speed	Spring vs. Fall	6.45 × 10−60
4	Average speed	Spring vs. Summer	3.92 × 10−23
5	Average speed	Winter vs. Fall	1.17 × 10−50
6	Consumption overall	Spring vs. Fall	2.21 × 10−85
7	Consumption overall	Spring vs. Summer	0.00 × 100
8	Consumption overall	Winter vs. Fall	9.93 × 10−307
9	Consumption overall	Winter vs. Summer	0.00 × 100
10	Consumption overall	Winter vs. Spring	4.33 × 10−125
11	Consumption overall	Summer vs. Fall	0.00 × 100
12	Distance driven	Winter vs. Spring	6.77 × 10−3
13	Distance driven	Winter vs. Fall	3.64 × 10−56
14	Distance driven	Spring vs. Summer	1.80 × 10−5
15	Distance driven	Spring vs. Fall	3.78 × 10−40
16	Distance driven	Summer vs. Fall	1.36 × 10−13
17	Distance driven	Winter vs. Summer	4.43 × 10−13
18	Energy regenerated driving	Spring vs. Fall	7.57 × 10−140
19	Energy regenerated driving	Spring vs. Summer	1.80 × 10−208
20	Energy regenerated driving	Winter vs. Fall	3.52 × 10−157
21	Energy regenerated driving	Winter vs. Summer	6.21 × 10−231
22	Energy regenerated driving	Summer vs. Fall	9.58 × 10−6
23	Energy used	Winter vs. Spring	2.94 × 10−25
24	Energy used	Summer vs. Fall	7.53 × 10−224
25	Energy used	Spring vs. Summer	0.00 × 100
26	Energy used	Winter vs. Fall	6.32 × 10−14
27	Energy used	Winter vs. Summer	0.00 × 100
28	SOC used	Winter vs. Spring	2.88 × 10−24
29	SOC used	Winter vs. Summer	0.00 × 100
30	SOC used	Winter vs. Fall	3.80 × 10−7
31	SOC used	Spring vs. Summer	0.00 × 100
32	SOC used	Summer vs. Fall	1.11 × 10−272
33	Time idling	Summer vs. Fall	7.67 × 10−77
34	Time idling	Spring vs. Fall	2.97 × 10−129
35	Time idling	Spring vs. Summer	4.74 × 10−2
36	Time idling	Winter vs. Summer	9.28 × 10−15
37	Time idling	Winter vs. Spring	1.28 × 10−33
38	Time idling	Winter vs. Fall	1.01 × 10−37
39	Time in service	Spring vs. Fall	3.85 × 10−3
40	Time in service	Winter vs. Spring	1.49 × 10−30
41	Time in service	Winter vs. Summer	2.72 × 10−74
42	Time in service	Winter vs. Fall	3.15 × 10−33
43	Time in service	Spring vs. Summer	7.81 × 10−40
44	Time in service	Summer vs. Fall	1.87 × 10−10

.

With the energy used and consumption overall, significant contrasts are observed across all seasonal combinations, with the most pronounced differences observed between winter and summer, as well as spring and summer, reflecting increased HVAC load and operational intensity during warmer months. The energy regenerated during driving is notably lower in the summer compared to spring and winter, suggesting reduced regenerative braking due to smoother traffic flow or fewer stop events. Average speed varies significantly across the seasons, peaking in spring and fall, likely associated with favorable road conditions and lighter congestion. Distance driven also differs across all seasonal pairs, particularly between winter and fall, indicating service expansion or adjusted schedules in response to seasonal ridership demand. The SOC used mirrors the consumption trends, with summer exhibiting the greatest battery draw due to extended usage and higher cooling needs. Time idling is significantly longer in winter and fall, likely due to longer warm-up durations or schedule slack, while spring shows the lowest idle times. Time in service, although more consistent, still indicates that winter has the longest operational periods—possibly due to delays or extended trip durations—making it clear that both vehicle usage and performance are seasonally sensitive and must be managed accordingly.

For post hoc comparisons, the rank-biserial correlation (r) has been used as the effect size for the Mann–Whitney U tests. This metric reflects the strength and direction of differences between two groups. It is calculated as follows:

$$r=1-{\displaystyle \frac{2U}{n_1\times n_2}}$$

where

• $U$ is the Mann–Whitney U statistic;
• $n_1$ and $n_2$ are the sample sizes of the two comparison groups.

In the summer vs. winter comparison for HVAC proxy energy, we obtained a U = 134,933, with n₁ = 1005 (winter) and n₂ = 1028 (summer):

$$r=1-{\displaystyle \frac{2\times 134933}{1005\times 1028}}\approx -0.745$$

The value of r = −0.745 indicates a very large effect size, further reinforcing the substantial seasonal impact on energy usage.

4.7. HVAC Proxy Energy Estimation and Validation

Methodology and Proxy Formula

In the absence of direct electrical telemetry, such as compressor torque or high-resolution HVAC power draw, HVAC energy consumption was estimated using a proxy formula based on secondary telematics signals. This method is widely adopted in academic research and commercial fleet monitoring applications when detailed subsystem-level data are unavailable [68]. The formula used is as follows:

HVAC_Proxy_Energy_kWh = (0.4 × Condenser Fan Speed_normalized + 0.4 × Cabin Fan Speed_normalized + 0.2 × Discharge Pressure_normalized) × 5

Each term in this formula represents a normalized signal associated with HVAC system activity. The fan speed components (cabin and condenser) are directly related to airflow demands and operate during both cooling and heating cycles. The discharge pressure serves as an indicator of compressor load and thermal system effort. The weights (0.4, 0.4, and 0.2) reflect estimated proportional contributions of condenser fan speed, cabin fan speed, and discharge pressure to total HVAC system effort. These were chosen based on component-level behavior reported in HVAC energy allocation studies for commercial EVs and electric buses [69]. The final scaling factor of five was derived through iterative adjustment to align monthly proxy energy totals with values reported in manufacturer specifications for Proterra EBs and in peer-reviewed studies of HVAC energy intensity in similar climates. Once ground-truth telemetry becomes available, a detailed sensitivity test on this multiplier is planned for future work.

Figure 11 shows the sensitivity analysis of the HVAC proxy energy estimation model under varying weighting schemes. Despite variations in weights assigned to fan speeds and discharge pressure, the model shows a high correlation with the baseline proxy, suggesting robustness of the estimation method.

Applying the proposed formula, the monthly average HVAC energy consumption was calculated and visualized in the time series plot presented in Figure 12.

This proxy-based estimation reveals a distinct seasonal variation in HVAC demand across the year. Summer exhibits the highest energy usage, reflecting intensified cooling loads due to elevated ambient temperatures. At the same time, winter shows the lowest, likely because heating requirements in electric buses are often less energy-intensive or supplemented by auxiliary heating systems.

Figure 13 presents the average HVAC energy consumption over a 24 h operational cycle for each season to investigate hourly variations further.

The results clearly indicate a midday peak, corresponding to rising external temperatures and passenger loads. Summer shows the most pronounced spike, followed by the fall and spring, both of which also reflect noticeable HVAC engagement. In contrast, winter maintains a relatively flat and low energy profile, reinforcing the notion of reduced HVAC dependency during colder months.

To assess the reliability of this proxy metric, it was compared with a simulated total vehicle energy consumption dataset. While actual total energy data was unavailable, a realistic energy profile was created by scaling the proxy HVAC values with stochastic noise to mimic operational variability, as shown in Figure 14.

A correlation analysis revealed strong associations between the proxy and simulated total energy, with a Spearman correlation of 0.897 and a Pearson correlation of 0.882. The black line represents a linear regression fit, highlighting the strong positive correlation, as shown in Figure 14. These results indicate that the proxy not only preserves the relative trends of HVAC energy use but also mirrors broader energy dynamics in the fleet. Such validation supports its use in operational studies where subsystem-level impacts, like HVAC load due to seasonal variation, need to be analyzed.

4.8. HVAC Optimization

The primary objective of this optimization is to minimize the total energy consumption of the HVAC system in electric buses while maintaining passenger comfort. The proposed model incorporates dynamic environmental and operational variables—specifically ambient temperature and passenger load—significantly influencing HVAC energy demand throughout the day. This optimization framework assumes that ambient temperature and passenger load data can be accessed or estimated in real time. In practice, temperature inputs can be collected using on-board sensors or third-party weather APIs, while passenger load is estimated using Automated Passenger Counters (APCs), infrared sensors, or farebox boarding data. These variables are normalized and used to adjust hourly HVAC demand throughout the day.

4.8.1. Methodology and Mathematical Formulation

The day is divided into 24 hourly intervals, indexed by $i\in \{0,1,\dots ,23\}$. For each hour, the following parameters are defined:

• $E_i^{\mathrm{base}}$: Baseline HVAC energy consumption (in kWh), based on nominal system usage.
• $T_i$: Ambient temperature (in °C).
• $P_i$: Passenger load (number of passengers).
• $f_T\left(T_i\right)$, $f_P\left(P_i\right)$: Normalized temperature and passenger load scaling factors.
• $E_i^{\mathrm{adj}}=E_i^{\mathrm{base}}\cdot f_T\left(T_i\right)\cdot f_P\left(P_i\right)$: Adjusted baseline energy consumption.
• $x_i$: Decision variable representing the actual HVAC energy consumption during hour $i$, to be optimized.

The ambient temperature and passenger load profiles used in this study were synthesized to mimic typical patterns observed in Central Florida’s urban transit operations. A sine curve centered around 20 °C simulates daily temperature variation, peaking near 2 p.m. Passenger load is approximated heuristically, reflecting peak ridership during morning and evening commute windows (6–9 a.m. and 4–7 p.m.) and lighter loads otherwise. This aligns with ridership patterns observed in LYNX services across high-frequency urban corridors.

• Objective Function

The goal is to minimize the total HVAC energy usage across all 24 h of the day:

$$\mathrm{m}\mathrm{i}\mathrm{n}\sum _{i=0}^{23}{x}_i$$

• Constraints

(i)

Comfort Constraint:

To maintain passenger comfort, energy consumption must stay within a reasonable band (60% to 100%) of the adjusted baseline:

$$0.6\cdot E_i^{\mathrm{adj}}\le x_i\le E_i^{\mathrm{adj}} \forall i$$

(ii)

Peak Load Constraint (12 p.m.–5 p.m.):

HVAC energy usage during peak demand hours (12:00 p.m. to 5:00 p.m.) should not exceed the average consumption during the off-peak hours:

$${\displaystyle \frac{1}{6}}\sum _{i=12}^{17}{x}_i\le {\displaystyle \frac{1}{18}}\sum _{i\notin \{12,13,14,15,16,17\}}{x}_i$$

(iii)

Idle-Hour Constraint (Scenario B, 1 a.m.–5 a.m.):

In a secondary scenario that emphasizes energy savings during non-operational hours, HVAC energy consumption is fixed at the lower limit (60% of adjusted baseline) between 1:00 a.m. and 5:00 a.m.:

$$x_i=0.6\cdot E_i^{\mathrm{adj}} \forall i\in \{1,2,3,4,5\}$$

The comfort constraint range (60–100%) of adjusted baseline HVAC energy usage was selected to reflect practical control flexibility in transit HVAC operations, especially under constrained energy budgets. While standards such as ASHRAE Standard 55 [70] and EN ISO 7730 [71] do not prescribe specific percentages of HVAC energy consumption, they emphasize maintaining thermal comfort through acceptable ranges of temperature, humidity, and air movement. The chosen lower bound of 60% serves as a conservative approximation to maintain minimal airflow and thermal conditioning during idle or low-occupancy hours, avoiding under-conditioning [72]. Although empirical studies validating this exact range are limited, the constraint is aligned with energy-aware HVAC management strategies reported in transit energy optimization research. Future work will aim to calibrate these thresholds using actual passenger comfort feedback and operational telemetry.

• Inputs

• Baseline Energy Consumption: Randomized between 2.5 and 5.5 kWh per hour.
• Temperature Profile: Modeled using a sine function centered around 20 °C, peaking near 2 p.m. to mimic diurnal variation.
• Passenger Load: Higher during commute hours (6–9 a.m. and 4–7 p.m.).

Both temperature and passenger load are normalized and used as multiplicative factors to adjust baseline HVAC energy demand dynamically.

4.8.2. Results

Figure 15 illustrates the hourly HVAC energy usage across three scenarios: the adjusted baseline (unoptimized), the optimized output based on ambient temperature and passenger load, and the reduction achieved by enforcing idle-hour constraints.

The optimized curve consistently lies below the baseline across all hours, with the most substantial savings observed during peak ambient temperature and passenger load periods. Notably, between 12 p.m. and 5 p.m., energy usage is reduced by the maximum allowable 40%, demonstrating the model’s ability to respond effectively to high-demand conditions while adhering to comfort constraints. Early morning and late-night hours show moderate reductions (28–30%), reflecting a reduced system load due to lower occupancy and milder temperatures. The visualized energy gap between baseline and optimized usage clearly highlights the system’s ability to maintain comfort while minimizing energy waste.

Figure 16 presents a comparative bar chart of total daily HVAC energy consumption under three different scenarios: original (no optimization), optimized, and optimized with idle-hour reduction.

The original scenario consumes 92.2 kWh, while the optimized scenario reduces this to 74.8 kWh—an 18.9% decrease. When idle-hour consumption is fixed at 60% of the adjusted baseline between 1 a.m. and 5 a.m., the total energy drops further to 70.1 kWh, achieving a 23.9% reduction. These results demonstrate that incorporating contextual inputs like temperature and occupancy already provide significant savings; however, modest operational policies, such as scheduling idle-hour minimization, can compound the efficiency gains.

4.8.3. Sensitivity Analysis

To test the robustness of the HVAC optimization model against input variability, a sensitivity analysis was performed by perturbing the ambient temperature and passenger load values by ±10%, both independently and simultaneously, as shown in

Table 4. Sensitivity analysis of optimized HVAC energy use under ±10% perturbations in input variables.

Scenario	Total Energy (kWh)	% Change from Optimized
Base	55.40	0.00%
+10% Temp	58.00	+4.71%
−10% Temp	52.79	−4.71%
+10% Load	57.76	+4.26%
−10% Load	53.04	−4.26%
+10% Both	60.48	+9.17%
−10% Both	50.54	−8.77%

.

The total optimized daily HVAC energy consumption varied by approximately ±4.7% when only temperature was altered and ±4.3% when only passenger load was perturbed. When both parameters were increased or decreased simultaneously by 10%, the variation reached +9.2% and −8.8%, respectively. These results confirm that the optimization model remains stable under moderate uncertainty in sensor inputs, affirming its suitability for real-world deployment where exact real-time telemetry may involve small errors.

5. Conclusions

This study uses over three years of high-resolution telematics data to deliver a detailed analysis of electric bus (EB) performance within the LYNX LYMMO fleet in Orlando, Florida. The analysis captured seasonal, temporal, and operational variations across key performance metrics, including energy consumption, regenerative braking, driving behavior, and HVAC system load.

It is also observed that idling time consistently exceeds driving time, marking it as the most impactful and controllable variable for improving energy efficiency. While energy consumption is affected by several external factors—including road gradient, slope, bus age, vehicle type, passenger load, and weather conditions—idling time is more amenable to operational control through better scheduling and HVAC integration. This insight underpinned the optimization model, which demonstrated that minimizing idling during low-demand periods can substantially reduce energy consumption without compromising passenger comfort. Notably, although the fleet exhibits relatively consistent driving distances and patterns across seasons, idling remains the key variable that can be actively reduced through operational strategies.

Furthermore, the integration of a validated HVAC proxy model enabled meaningful subsystem-level insights despite the absence of direct telemetry. A dynamic optimization model incorporating ambient temperature and passenger load demonstrated the potential to reduce HVAC energy consumption by up to 24%. Overall, the findings highlight the importance of data-driven energy management strategies in maximizing electric transit fleets’ operational efficiency and sustainability.

6. Discussions

This study highlights the influence of seasonal dynamics and operational factors on the energy efficiency of electric transit fleets. Among all variables, idling time emerged as the most impactful and operationally controllable contributor to excess energy consumption, offering opportunities for significant efficiency gains without new infrastructure.

Unlike average speed or regenerative braking—which showed moderate seasonal variation—idling time fluctuated more substantially and strongly correlated with HVAC energy use. In winter, idling was associated with layovers and vehicle warm-up, while in summer, excessive HVAC cooling during idling substantially increased total energy demand. Addressing idling, therefore, is critical not only for propulsion energy savings but also for auxiliary subsystems.

A key finding of this study is that the summer, not winter, was the peak season for energy consumption, contrary to most of the literature from cold climates where winter dominates due to heating loads and battery inefficiencies. In Florida’s subtropical environment, high HVAC cooling demand drove this shift. Winter, conversely, showed the lowest energy use, marking a novel contribution to the electric bus performance literature.

Additionally, HVAC systems accounted for a large share of total energy use, particularly during idling. This finding challenges the traditional assumption that propulsion is the dominant energy consumer. Targeting HVAC operation during non-driving periods offers a practical and scalable strategy for reducing energy use.

Based on the normalized ambient temperature and passenger load, the HVAC proxy model proved effective in estimating subsystem energy use. Although lacking direct telemetry validation, its strong correlation with simulated trends supports its application in data-limited contexts. A ±10% sensitivity analysis of input noise showed minimal effect (<9.2%) on optimized energy use, confirming the model’s robustness. However, because the proxy and benchmark share functional inputs, there is a potential for circularity. Future work will integrate manufacturer specifications and on-board telemetry to improve model validation.

This study focused on HVAC optimization due to its significant share of idle energy use and controllability via real-time ambient and occupancy data. Other subsystems, such as battery thermal management and drivetrain behavior, were excluded due to data limitations but are targets for future research.

High ambient temperatures may also affect battery performance. LiFePO₄ batteries used in the LYMMO fleet can experience thermal derating, voltage drop, or capacity fade during rapid charging or prolonged exposure to heat. While our analysis centered on operational HVAC energy, future work should integrate battery thermal performance modeling to capture seasonal impacts fully.

Compared to findings from colder climates like Canada, Finland, and Germany—where winter is typically the most energy-intensive—our results demonstrate that in hot, humid cities like Orlando, summer imposes the most significant energy burden. Preliminary data from cities such as Houston, Miami, and Bangkok suggest similar patterns, reinforcing the need for climate-sensitive optimization strategies.

As agencies move from pilot programs to full-scale electric fleet adoption, prioritizing operational inefficiencies such as HVAC usage during idling can yield rapid, low-cost improvements in system-wide efficiency. The insights presented here directly apply to the LYNX LYMMO fleet and are transferable to similar urban electric transit systems operating in subtropical and tropical climates.

7. Recommendations and Limitations

This study recommends prioritizing idling reduction and HVAC energy optimization to enhance electric bus efficiency. Since idling often exceeds active driving time and drives substantial energy waste during HVAC operation, agencies should adjust schedules to minimize idle periods and adopt smart HVAC control strategies based on ambient temperature and passenger load. The validated HVAC proxy and optimization model demonstrated up to 24% energy savings and can be integrated into real-time fleet management systems.

Additionally, agencies operating in warmer climates should consider region-specific deployment strategies. In contrast to the existing literature which identifies winter as the peak season, this study highlights summer as the highest energy use period in Florida, largely due to HVAC cooling loads. Future work should extend the optimization model to include propulsion and battery thermal management and test its adaptability across different climate zones.

However, the study has limitations. HVAC energy use was estimated through proxies, not direct telemetry, introducing some uncertainty. The model shows strong internal consistency (Pearson = 0.882, Spearman = 0.897) but lacks independent validation. As the simulated total energy was derived from the proxy with added noise, this creates potential circularity. Future validation using direct HVAC telemetry is essential.

Passenger load was approximated using standard ridership patterns without real-time sensor data. Vehicle-level variations—such as bus age, driver behavior, and route topology—were not included, possibly affecting energy outcomes. These factors should be integrated into future models.

The analysis did not quantify temperature-related battery inefficiencies, which may increase summer energy use. Thermal modeling of LiFePO₄ battery performance should be included in future work.

Disaggregating idle energy consumption by source (e.g., layovers vs. traffic) using subsystem telemetry would support more targeted interventions. Distinguishing between controllable and uncontrollable idling sources can guide specific energy-saving strategies.

Because the findings are based on Central Florida data, results may not generalize to colder climates with different seasonal energy profiles. Regional HVAC behavior, congestion patterns, and charging needs should be considered when applying the model elsewhere.