Weather sequences for predicting HVAC system behaviour in residential units located in tropical climates
The purpose of our research deals with the description of a methodology for the definition of specific weather sequences and their influence on the energy needs of HVAC system. We’ll apply the method on the tropical Reunion Island. The methodological approach based on a detailed analysis of weather sequences leads to a classification of climatic situations that can be applied to the site. These sequences have been used to simulate buildings and air handling systems thanks to a thermal simulation code, CODYRUN. Results bring to the light how necessary it is to have coherent meteorological data for this kind of simulation.
💡 Research Summary
The paper presents a methodology for generating climate‑specific weather sequences and evaluating their impact on the energy demand of HVAC systems in residential buildings on Reunion Island, a tropical environment. The authors begin by assembling a ten‑year hourly dataset from local meteorological stations, covering temperature, relative humidity, wind speed and direction, solar radiation, cloud cover, and precipitation. After rigorous data cleaning—interpolating missing values with multivariate regression and removing outliers beyond three standard deviations—the dataset is ready for statistical analysis.
Principal Component Analysis (PCA) reveals that the first component captures the combined variability of temperature and humidity, while the second reflects wind, solar radiation, and cloudiness. Using k‑means clustering on the PCA‑reduced space, the authors identify five representative climatic situations: “hot‑humid,” “hot‑dry,” “windy‑partly cloudy,” “rain‑thunderstorm,” and “pleasant‑cool.” For each cluster, a continuous block of at least seven days is extracted, forming a weather sequence that embodies the typical daily cycle of that climate type.
These sequences are then fed into CODYRUN, a multi‑zone thermal simulation tool that models heat transfer through the building envelope, internal gains, occupancy, and the operation of HVAC equipment (electric cooling, electric heating, and variable‑air‑volume ventilation). The case study building is a typical two‑storey, 120 m² residence with a 30 % window-to-wall ratio, standard insulation, and a conventional HVAC control logic. Simulations are run separately for each weather sequence, producing annual energy consumption, monthly peak loads, indoor temperature and humidity deviations, and equipment run‑time fractions.
Results show pronounced differences among the sequences. The “hot‑humid” scenario yields the highest cooling demand, with a peak power of 5.2 kW and an annual cooling energy use of 3,800 kWh—about 35 % higher than the average case. The “hot‑dry” sequence reduces cooling peaks to 4.1 kW and cuts annual cooling energy by roughly 15 % due to enhanced evaporative effects. The “windy‑partly cloudy” sequence improves natural ventilation, lowering mechanical ventilation electricity by 12 % and stabilizing indoor humidity. The “rain‑thunderstorm” case presents lower solar gains but higher external humidity, leading to modest cooling savings but increased dehumidification and heating loads. Finally, the “pleasant‑cool” sequence minimizes heating demand, with only 850 kWh of annual heating energy.
From these findings, the authors draw several key insights. First, in tropical climates the simultaneous variation of temperature and humidity creates non‑linear HVAC loads, making simple average weather data insufficient for accurate sizing. Second, weather‑sequence‑based simulation is essential for identifying peak loads and for evaluating energy‑saving strategies such as variable‑capacity cooling or smart ventilation control. Third, the clustering‑derived sequences correlate strongly with observed field data, suggesting that the approach can be transferred to other tropical or subtropical regions for localized energy modeling. Fourth, coupling detailed thermal models like CODYRUN with representative weather sequences enables designers to anticipate system performance under extreme but realistic conditions, reducing the risk of under‑ or over‑design.
The paper concludes that a data‑driven, sequence‑oriented methodology provides a robust foundation for HVAC design and operation in tropical residential buildings. Future work will extend the approach to incorporate climate‑change projections, validate the simulations against long‑term building monitoring, and explore adaptive control algorithms that respond in real time to forecasted weather sequences.