Multiobjective optimisation of passive energy saving strategies in Zhuangke dwellings in Huangyuan county, China

Abstract

In rural areas of Huangyuan County, Qinghai Province, traditional dwellings are gradually being replaced by self-built brick-concrete houses that mimic urban designs, leading to increased energy consumption and reduced indoor thermal performance. This issue is particularly urgent in economically disadvantaged regions with harsh climatic conditions. Traditional Zhuangke dwellings, however, feature architectural characteristics that are well-suited to the cold climate and limited local resources. Extracting and optimising the passive energy-saving design of these traditional dwellings presents an effective solution to address the conflict between energy efficiency and indoor thermal performance in self-built rural houses. This study Innovatively develops a multi-objective optimisation framework for passive energy-saving design in rural housing in Huangyuan County, using the Rhino and Grasshopper visual programming platform. The framework aims to resolve this conflict. By analysing the design parameters of 53 traditional Zhuangke dwellings, the study optimises 10 passive energy-saving design parameters, including courtyard length, width, building orientation, height, depth, span, ground clearance, window-to-wall ratio, sunroom depth, and north-facing double-wall cavity depth. The Technique for Order Preference by Similarity to Ideal Solution​ (TOPSIS) method is then applied to filter the Pareto optimal solutions. The results show that, compared to the baseline building, the optimised “linear courtyard” Zhuangke dwelling achieves a reduction in heating energy use intensity (EUI) by 1–17.9% and an improvement in indoor standard effective temperature (SET) ranging from 4.1 to 5.4%. This study provides quantitative evidence and practical design strategies tailored to linear courtyard-type Zhuangke dwellings in rural Qinghai, contributing to the reduction of heating energy consumption and the improvement of indoor thermal performance in cold-climate self-built housing.

Introduction

Traditional Zhuangke dwellings represent the outcome of prolonged adaptation to harsh cold climates, scarce resources, and multicultural influences. These structures utilise spatial and interface designs through iterative refinement to achieve satisfactory indoor thermal performance with minimal energy consumption¹. However, field surveys reveal that with the rapid advancement of urbanisation, rural residences in Huangyuan County are predominantly constructed through self-building practices. There is a tendency to unthinkingly follow the brick-concrete residential model while neglecting energy efficiency considerations. This has led to increased heating energy consumption and suboptimal indoor temperatures during winter. Therefore, investigating and optimising traditional dwellings’ passive energy-saving design principles and integrating these into local residential designs represents an effective strategy to reduce winter heating energy consumption and enhance thermal performance in the region.

Numerous studies have investigated traditional rural dwellings’ passive energy-saving optimisation design using field surveys and simulation software². Under a performance-oriented approach, the research mainly focuses on optimising building envelope structures, solar air heat energy systems, and architectural space design. To optimise the building envelope structure, the standard model and data analysis were adopted to optimise the window-to-wall ratio and building orientation of residential buildings in the cold region of China regarding energy consumption and natural lighting³. The energy efficiency and indoor thermal comfort of the building envelope of residential buildings in Marrakesh, Morocco, were optimised by using the artificial neural network (ANN) and multi-objective particle swarm optimisation (MOPSO) algorithm⁴. Adopting on-site monitoring and Ecotect simulation has improved the indoor temperature fluctuations and energy efficiency of rural residential buildings in Yulin, China⁵. The design of building exterior walls in the low-temperature climate zone of China was optimised by adopting a multi-objective optimisation algorithm (NSGA-II) and EnergyPlus, aiming to enhance indoor thermal comfort and reduce life cycle cost, building energy consumption and CO₂ emissions^6,7. Combined with BO-XGBoost-NSGA-II, multi-objective optimisation theory was employed to optimise the envelopes of residential buildings in China’s hot-summer and cold-winter regions. The aim was to reduce energy consumption while improving thermal comfort, daylighting^8,9 construction time, costs, quality, and CO₂ emissions¹⁰. Multidisciplinary optimisation (MDO) and multi-objective simulations were applied to analyse structures in Pennsylvania, USA. The findings emphasised that early-stage design collaboration reduces embodied carbon and enhances efficiency¹¹. BIM¹² multi-objective optimisation methods and simulation platforms are employed to strengthen structural and environmental performance integration in the early architectural design^13,14. In optimising solar air thermal energy systems, numerical simulation and regression analysis were adopted to optimise the solar air heating system for preheating ventilation air in cold regions to enhance thermal storage efficiency¹⁵. The Hooke-Jeeves algorithm was adopted to optimise the solar air-source heat pump system in China’s high-altitude and cold areas, saving energy and maintaining indoor thermal comfort¹⁶. In terms of optimising the design of architectural spaces, the DesignBuilder software was adopted to simulate and optimise the heating energy efficiency of the traditional residences in the Tibetan areas of Sichuan Province regarding their orientations, architectural shapes and enclosure structures¹⁷. Optimisation and climate adaptability theories, coupled with Ecotect and Phoenics simulations, were utilised to examine the courtyard and roof designs of traditional courtyard houses in Xuzhou. The study demonstrated that optimising these features enhanced thermal performance and energy efficiency¹⁸. The impact of sunrooms on the indoor thermal conditions, energy efficiency, and economic performance of rural homes in southern Shandong Province, China, was assessed using Energy Plus software and dynamic payback period analysis¹⁹. The key design elements, including orientation, plan form, window-to-wall ratio, and roof slope, of traditional rural dwellings in the Qinba Mountain region were analysed using the Wallacei-X multi-objective optimisation algorithm²⁰.

Previous studies have shown that the characteristics of traditional passive design effectively respond to the local climate, providing a comfortable indoor thermal environment, low energy consumption, and significantly reducing total energy demand^21–23. While significant research has been conducted on passive energy-saving design for urban residences in cold or hot-humid climates in China^8,9,21,24,25 studies specifically targeting rural housing in economically underdeveloped and cold regions, such as northwest China, remain sparse. Existing studies have focused predominantly on Eastern China’s urban areas, overlooking the unique climatic and economic conditions in rural northwest regions like Huangyuan County.

Most studies on passive design optimisation have emphasised building envelope improvements, such as window-to-wall ratios, structure and materials^{3,5–7,10,13,20,21,26} and solar energy or air heat systems^15,16. However, these studies often neglect a comprehensive quantitative analysis of traditional housing form and spatial design parameters^17–19. Specifically, there is a lack of research on how optimising form and spatial design parameters (e.g., courtyard dimensions, building depth, height, and span) impacts building performance. Additionally, a holistic, multi-objective optimisation approach that addresses energy consumption, thermal comfort, and other factors (e.g., construction cost, CO₂ emissions, daylighting, and ventilation) for traditional rural dwellings is still underexplored. Most existing studies rely on single-factor optimisation methods, whereas a comprehensive multi-objective approach is required to balance multiple performance goals.

Moreover, research on the passive design characteristics of Zhuangke dwellings in Qinghai’s rural areas has been chiefly qualitative, without detailed quantitative simulations or evaluations^1,22. The existing literature lacks studies that comprehensively combine energy-saving measures and indoor thermal performance improvement in a multi-objective optimisation context for traditional rural housing in Qinghai. There is a substantial gap in applying advanced simulation tools like EnergyPlus to assess passive design strategies combined impact on energy efficiency and thermal performance. While there have been several studies using multi-objective optimisation algorithms like MOPSO, NSGA-II, and ANN to optimise residential building design in various regions^{4,6–9,13,17,18,27,29} their application to the specific case of Zhuangke dwellings in Huangyuan County, Qinghai, is missing. These methods have not yet been applied to study passive design’s combined impact on energy-saving and thermal performance in this context.

Most existing studies rely on single-factor optimisation methods, whereas a comprehensive multi-objective approach is required to balance multiple performance goals such as energy efficiency, thermal comfort, and construction cost. Current technical frameworks for multi-objective optimisation in passive energy-saving design can be categorised into three types: general programming platforms (e.g., Matlab, GenOpt, CAMOS)³⁹ hybrid programming-simulation platforms (e.g., Matlab-EnergyPlus, GenOpt-DOE-2)^4,30,40 and architectural visual programming platforms (e.g., Rhino + Grasshopper, Revit + Dynamo) ^41,42. However, the first two categories face challenges of high technical barriers and poor interactivity, limiting their adoption in architect-led design processes. Consequently, this study adopts Rhino + Grasshopper—a widely used visual programming tool in architecture—as the optimisation framework, integrated with Honeybee for building performance simulation and Octopus for multi-objective algorithm execution.

To address the identified gaps, this study selects seven villages in Huangyuan County, where traditional dwellings are well-preserved and concentrated, as case study locations. Data on the design parameters, winter heating energy structure, and usage patterns of traditional Zhuangke dwellings in these villages were collected through field investigations. Statistical methods and multi-objective optimisation algorithms were then applied to extract and optimise the passive energy-saving design parameters of Qinghai’s Zhuangke dwellings, which were summarised into design strategies. Based on the identified research gaps, this study aims to address the following three questions:

1. How can the passive energy-saving design parameters and their values be extracted from traditional Zhuangke dwellings?

2. How can we optimise these passive design parameters with the goals of energy conservation and indoor thermal performance?

3. How can strategies for guiding energy-saving design in residential buildings in Qinghai Province be summarised?

The results of this study are expected to contribute to developing a knowledge system for passive energy-saving design in residential buildings in Huangyuan County and the surrounding cold regions of Qinghai. Specifically, the study comprehensively considers heating energy consumption and thermal performance, summarising eight passive design strategies (including ten design parameters) from both building space and envelope perspectives. This research addresses the gap in quantitative analysis of passive energy-saving design for traditional Zhuangke dwellings. It provides a scientific basis for the design and optimisation of rural housing in Qinghai Province. Furthermore, it offers practical recommendations for improving building design in similar climatic environments.

Materials and methods

The research methodology encompasses field research and multi-objective optimisation algorithms. Field surveys were conducted to collect data on the form design parameters of Zhuangke dwellings. A foundational model of these residences was developed using Rhino and Grasshopper parametric modelling techniques. A parametric building performance simulation engine was then employed to model key performance indicators such as heating energy consumption and thermal performance. Multi-objective optimisation algorithms were applied to refine architectural design parameters, resulting in an optimised solution set that informed the development of residential design strategies. Figure 1 provides a visual representation of the research framework, enhancing the clarity of our approach.

Fig. 1.

Research Process and Research Methods.

Site and climate

Huangyuan County, located in Xining City, Qinghai Province (10054’ − 10125’ E, 3620’ − 3653’ N), lies on the northeastern fringe of the Qinghai-Tibet Plateau. It is famous for its intact traditional architecture and the concentration of villages throughout the region³¹. The county spans an area of 1,545 km², with altitudes ranging from 2,430 to 4,898 m. Notably, villages along Riyue Mountain and the Hongshui River offer concentrated distributions and excellent architectural samples for study. Considering village size, sample diversity, and traditional features, representative villages including Yixi, Old Tu’er Gan, New Tu’er Gan, Xiao Gaoling, Chaqu, Cha Hansu, and Men Gudao were chosen for their varying sizes, traditional architectural forms, and well-preserved courtyard houses, many of which have undergone renovations or reconstructions in recent decades. Figure 2 shows the location map of the surveyed area.

Fig. 2.

Map of Huangyuan County, Qinghai Province, China.

Huangyuan County has a cold semi-arid climate (BSK) according to the Köppen-Geiger classification, which is characterised by relatively low yearly temperatures and considerable solar radiation³². Climate data from the most recent Typical Meteorological Year (TMY) database³³ indicate severe winters (from December to February), with temperatures dropping to −13 °C and an average monthly range spanning from − 15 °C to 1 °C. Despite these cold conditions, the area enjoys substantial solar radiation, receiving an annual total between 571.67 and 701.66 W/m², alongside around 2,719 h of sunlight annually. The meteorological data of Huangyuan County are shown in Fig. 3 above.

Fig. 3.

Comprehensive map of typical annual temperature radiation in Huang Yuan County.

Data collection and screening

Zhuangke dwellings, a typical residential style in Qinghai Province, are characterised by architectural elements that respond to the local climate³⁴. Primarily located in eastern Qinghai near the Yellow and Hapu Rivers, these homes are constructed using locally available materials and straightforward building techniques, allowing easy assembly and providing strong thermal performance. The design of these dwellings ensures that they retain heat during the winter, stay cool in the summer, and enhance energy efficiency throughout the year.

Considering the vast heating energy demand of Zhuangke dwellings in winter (from November to February)²²indoor occupancy rates and demand for thermal comfort are high during winter. The data collection for this study will be conducted in the coldest month when the heating demand is at its peak (January 27th to February 7th, 2024). The collected data will cover the design parameters of the courtyard houses’ form, the energy structure for heating, the residents’ basic information, and the houses’ usage situation.

Sampling survey of Zhuangke dwellings

Huangyuan County, Qinghai Province, has jurisdiction over seven townships, two towns, and 146 administrative villages. Among them, Riyue Tibetan Township and Heping Township cover the most significant area (41.3%), with concentrated villages. Therefore, this study selected seven villages under the jurisdiction of these two townships, which have substantial differences in village scale and relatively well-preserved residential buildings, to conduct systematic field mapping work. The formula for calculating the sample size of representative residential buildings (see Eq. (1)) was used for the stratified sampling of the above villages. With a degree of freedom of 1, a confidence level of 0.75, and a statistical error not exceeding 15%, the total sample size was calculated to be 53 households. The sample size S of each village is shown in Table 1. The basic information on the villages surveyed in Huangyuan County is as follows.

1

Table 1.

The number of village samples and the distribution of rural residences.

Township	Rizhao Tibetan autonomous township			Heping township
village	Yixi	Old Tu’ergan	New Tu’ergan	Xiao Gaoling	Chaqu	Cha Hansu	Men Gudao
Household number	7	10	7	11	5	6	7

In the formula, S represents the recommended sample size individually. is the chi-square value representing one degree of freedom under the preset reliability; N is the total sample size; ME is the design error range, %; p is the population proportion.

Field research reveals that the architectural design of Zhuangke dwellings can be categorised into several types, including the “linear courtyard house,” “L-shaped courtyard house,” “U-shaped courtyard house,” and “closed courtyard house“¹. Among them, the number of “linear courtyard houses” is the largest. The main residential sample buildings in Yixi Village and Old Tu’ergan Village were constructed in the 1980 s and renovated (mainly with the addition of sunrooms) after 2000. The residential samples in Xiao Gaoling Village, Cha Qu Village, Cha Hansu Village and Mongolian Dao Village were mainly renovated between 2008 and 2020. The structural types of the Zhuangke dwellings in Zhuangke Village are primarily brick and wood, followed by earth and wood structures (constructed around 1980). Winter heating energy consumption mainly relies on electricity and wood. The family structure is mostly three generations, and the majority have a permanent population of 4 or more. The basic information on the villages and dwellings surveyed is shown in Fig. 4.

Fig. 4.

Basic Information of the Surveyed Villages.

Selection of design parameters for passive energy-saving

⑴ Determine design parameters for passive energy-saving residential building forms.

International green building standards, such as ISO 5200, ASHRAE 90.1, Passive House, and LEED, emphasise the impact of design factors like shape coefficient (Surface-to-Volume Ratio), orientation, window-to-w.

all ratio, envelope performance, ventilation, shading, architectural form, and compactness on energy demand and efficiency^7,18,35,36. This study qualitatively identified and extracted these key parameters based on the above standards and classified them into spatial and interface levels. Figure 4 shows the passive energy-saving parameters of residences determined by this study.

⑵ Screen and statistically analyse the collected data.

Based on the measured data of the design parameters of 53 courtyard houses, the study adopts the Box Plot method to identify and remove outliers in the database. It determines the reasonable range of values and the median of each design parameter according to the Box Plot’s upper and lower quartile ranges³⁷. It also ensures that the design parameters are within the limits of rural economic and technical constraints, construction habits, lighting and energy-saving functions requirements, and rural residential design specifications.

Data analysis

The issues considered in architectural design are often diverse and comprehensive. Usually, there are contradictions and trade-offs among different goals, and an optimal solution set that maximally satisfies multiple goals needs to be sought. Multi-objective optimisation can consider the interactions and trade-offs among various objective functions simultaneously, and its optimisation results are usually represented as a set of Pareto front optimal solutions^4,38.

In this study, a visual programming platform commonly used in architectural practice—Rhino, combined with Grasshopper—was adopted to implement the optimisation framework. Honeybee was employed as the building performance simulation engine, and Octopus was the carrier for the optimisation algorithm. Based on field survey data, a multi-objective optimisation framework was developed to support the passive energy-saving design of traditional Zhuangke dwellings in Qinghai. The workflow, illustrated in Fig. 5, consists of the following four components:

Fig. 5.

Technical framework for multi-objective optimisation of passive energy-saving design in Zhuangke.

1. Generation of benchmark building models.

2. The setting of optimisation objective functions.

3. Optimisation solution sets and passive energy-saving design strategies.

4. Optimal solution decision-making.

Generation of benchmark Building models

Based on field research, the “Linear Courtyard House” Zhuangke dwellings represent the most significant proportion of the sample. Given the article’s space constraints, the study selects a representative “Linear Courtyard House” as the basis for constructing the core model. The layout of a “linear courtyard” Zhuangke dwellings typically comprises walls, a central courtyard, gates, and various structures (including the main house, side rooms, and corner rooms). The main house is conventionally situated on the north side with a sunroom, separated from auxiliary spaces such as kitchens and livestock sheds. The courtyard often features flower beds and a central area, which is usually aligned with the Buddha Hall. The gate is strategically positioned in one of the corners^1,22. The plan and three-dimensional model of a typical"linear courtyard” Zhuangke dwelling are shown in Fig. 6. The architectural floor plan was developed using AutoCAD 2020, and the 3D model was constructed in SketchUp Pro 2021 to support geometric and spatial analysis.

Fig. 6.

Plan and three-dimensional model of a typical “linear courtyard” Zhuangke dwelling. (a) Architectural floor plan created using AutoCAD 2020 (Autodesk Inc.; https://www.autodesk.com); (b) 3D building model constructed using SketchUp Pro 2021 (Trimble Inc.; https://www.sketchup.com).

(1)Benchmark architectural design parameters.

A parametric model concerning the form was constructed based on the typical “Linear Courtyard” Zhuangke dwelling in Fig. 5. Since the newly built rural residences in Huangyuan County after 2000 were all rebuilt on the original sites without any new land allocation, and to better represent the typical situation of the 53 households and avoid the interference of extreme values to the model, this study took the median of the survey data as the initial design parameter for the reference building. The GH model is shown in Fig. 7 (a).

Fig. 7.

Parametric model (a) and Structural Optimisation Model (b) of “Linear Courtyard” Zhuangke dwelling. Created using Ladybug Tools (v1.4.0; https://www.ladybug.tools).

⑵ Standard construction practice of buildings.

Regarding the setting of construction parameters, after investigation, it was found that the main structure of traditional village courtyards in Huanyuan County, Qinghai Province, is mainly brick and wood structures built by conventional techniques. The construction methods and thermal coefficients were determined based on the typical one-way courtyard residential buildings and referring to the construction design, combined with GB/T 50,824 − 2013 “Energy-saving Design Standard for Rural Residential Buildings“⁴³ and measured data. The construction methods of building envelope structure and the setting of heat transfer coefficients are shown in Table 2. The environmental analysis parametric plugin Honeybee (HB-Energy) of R&G was used to define the parameters (materials, structures, and thermal engineering parameters) of the building envelope structure to simulate energy and thermal performance⁴³.

Table 2.

The structure of typical Zhuangke dwellings.

Component	Construction (from the inside out/from top to bottom)	Thermal transmittance W/(m²·K)
Courtyard walls	440 mm rammed earth + 30 mm grass mud mortar + 30 mm original soil mortar	1.01
Interior wall	220 mm Red brick blocks + 15 mm lime plastering	0.89
Roof	60 mm Purlin + 40 mm Plywood board + 10 mm Tile + 150 mm Loess + 50 mm Silt layer	0.337
Pedestal base	450 mm Stone material	0.92
Window/ Inner door	Broken Bridge Aluminium Single-Layer Glass (6 mm)	5.8
Outer door	50 mm Plywood board	2.51

Figure 7 (b) shows the model with the setting of architectural structure and material parameters completed.

⑶ Baseline building operation information.

The configuration of building operation parameters encompasses energy consumption types, personnel schedules, heating temperature and duration, and other load-related information. Energy consumption types are customised based on conditions derived from field surveys and interviews using the HB program type function within the HB-Energy plugin suite. Personnel activity schedules, including occupancy density and indoor presence rates, are defined hourly using the HB people function. Indoor heating setpoints and operational periods are established using the HB setpoint function. Regarding heating systems, intermittent heating schedules were adopted to simulate the actual behaviour observed in rural households, where heating is typically limited to two hours in the morning and evening. This reflects local practices aimed at reducing fuel consumption and minimising expenses. Heating setpoints were defined using the HB Setpoint function, with indoor temperatures during heating periods constrained to a relatively low range (typically 14–16 °C), in order to minimise external energy inputs and isolate the passive thermal performance of the building envelope. Equipment operation, lighting, ventilation, and airtightness parameters adhere to GB/T 50,824 − 2013 “Energy-saving Design Standard for Rural Residential Buildings“⁴⁴ and JGJ 26-2018 “Energy-saving Design Standard for Cold Regions Residential Buildings“⁴⁵.EnergyPlus, as the core simulation engine integrated within the Honeybee platform, was employed to evaluate the building’s energy performance and indoor thermal comfort. Climatic inputs were obtained from the China Standard Meteorological Database (CSMD), and simulations were conducted for the coldest week of the TMY (February 17–23), during which outdoor temperatures ranged from − 6.86 °C to 3.29 °C, to generate SET results. In addition, the model computed the annual EUI (kWh/m²·year) and heating energy consumption per unit area during the winter heating season.

Multi-objective optimisation and goal setting

⑴ Building performance simulation indicators.

This study’s energy consumption and thermal performance evaluation indicators refer to the narrow definition of building operational energy consumption⁷. The building energy consumption evaluation indicators include EUI, cooling and heating loads, Renewable Energy Ratio (RER), etc^9,25,46. The indicators for evaluating the thermal performance of buildings include SET, Predicted Mean Vote (PMV), Relative Humidity (RH), etc^3,4,47,48. EUI, adopted by ISO 52,000, LEED, and ASHRAE 90.1, serves as a standardised metric for building energy consumption (kWh/m²/year), ensuring cross-regional comparability. ISO 52,000 designates EUI as a key indicator for operational energy accounting, LEED uses it as a baseline for energy-saving targets, and ASHRAE 90.1 applies it to evaluate compliance with building energy efficiency standards, particularly supporting energy reduction through passive design strategies. Meanwhile, SET is employed to evaluate thermal comfort under highly dynamic indoor conditions. Unlike steady-state models such as PMV, SET integrates multiple parameters—air temperature, mean radiant temperature, humidity, air velocity, metabolic rate, and clothing insulation—and captures the cumulative thermal effects of intermittent heating, radiant asymmetry, and localised discomfort risks from cold surfaces, making it highly suitable for non-uniform environments characteristic of passive design. It also allows for climate-specific adjustments, such as de-emphasising humidity and increasing insulation values, aligning with Qinghai’s dry, solar-dominated cold climate. SET is recognised by ASHRAE 55 and ISO 7730 for non-steady-state comfort analysis and has been validated through international certification systems such as LEED and WELL. The combined use of EUI and SET provides a robust, standardised framework for optimising energy performance and thermal comfort in passive buildings under extreme climatic conditions.

a. Heating energy use intensity (EUI).

The benchmark model’s winter heating energy consumption was simulated by calling EnergyPlus through Honeybee. The standard EPW files of Huangyuan County from the China Standard Meteorological Database (CSMD) were imported into the simulation. The building thermal condition parameters are shown in Table 2. According to the simulation results, the calculation formula for the heating energy consumption per unit building area is:

2

In the formula, Q_c is the heating energy consumption per building area unit. Q_ci represents the heating energy consumption of each room, measured in kW·h, and A_i indicates the area of each room, measured in m².

b. Standard effective temperature (SET).

The winter heating energy consumption of the benchmark model was simulated through the HB PMV Comfort Map in the HB-Energy plugin group, with meteorological data being incorporated and parameters such as calculation load and human activity level being set. Based on the simulation results, the general formulation can be expressed as:

3

In the formula: T_a represents the air temperature (°C), P_wv is the partial pressure of water vapour (kPa), v is the airflow velocity (m/s), 0.393 is the empirical coefficient for the influence of water vapour partial pressure on temperature, 0.027 is the empirical coefficient for the impact of wind speed on temperature, and 37 (°C) is the approximate value of the core body temperature of the human being.

⑵ Multi-objective algorithm.

The Honeybee computational engine can directly invoke the EnergyPlus calculation core to complete the simulation analysis of the building’s annual 8760-hour cooling and heating loads, dynamic energy consumption, etc. The SPEA2 (Strength Pareto Evolutionary Algorithm 2) built into the Octopus computational engine is compared with other multi-objective optimisation algorithms. This algorithm optimises the solution set based on the mechanism of natural selection, avoiding optimal local solutions and maintaining the uniform distribution of solutions. It can be better to avoid falling into local optima⁴⁹. Its fitness function is:

4

In the formula, R(i) represents the individual dominance information of individual i in the external and evolving populations; D(i) indicates the degree of crowding between individual i and its immediate neighbouring k-th individual.

This paper uses the optimisation objectives for the winter heating EUI and SET. The mathematical expression of the multi-objective optimisation problem is as follows:

5

The formula represents the heating energy consumption per unit building area (kW·h)/m²; represents the indoor standard effective temperature, h; represents the optimisation variable parameters.

Optimal solutions and passive energy-saving strategies

In multi-objective optimisation algorithms, multiple Pareto-optimal solutions are generated through the optimisation algorithm. The Pareto front can help decision-makers understand the trade-off relationship between energy efficiency and thermal performance objectives, enabling them to make the best design decisions based on actual requirements^4,6,10. Based on the optimisation solution sets of each design parameter, strategies for guiding the design are summarised from both the spatial and interface perspectives, including the range and tendency of parameter values.

Optimal decision-making

However, determining the final solution among these feasible optimal schemes requires decision-making. The TOPSIS comprehensive evaluation method is a practical decision-making approach which ranks the closeness of the finite number of evaluated objects to the idealised goal⁵⁰also known as the superiority and inferiority distance method. TOPSIS employs the distances between each solution and the positive ideal solution and the negative ideal solution to calculate the degree of proximity of each solution. This method evaluates the optimisation results and thereby screens out the optimal solution, determining the best scheme⁵¹.

The calculation formula is:

6

In the formula, and represent the distances of each evaluation solution to the positive and negative ideal solutions, respectively; C_j is the j-th evaluation solution, where 0 ≤ C_j ≤ 1. The closer C_j is to 1, the more optimal the evaluation solution becomes.

Results

Design parameters of Zhuangke dwellings

The design parameters influencing passive energy conservation in traditional courtyard houses include courtyard length, courtyard width, building orientation, building height, building depth, building span, height above the ground, window-to-wall ratio, sunroom depth, and the depth of the northern double-layer wall cavity (both spatial and interface categories). Figure 8; Table 4 show each passive energy-saving design parameter’s categories, names, forms, value ranges, and box-type analysis diagrams.

Fig. 8.

Passive energy-saving design parameters for Zhuangke dwellings.

Table 4.

Optimal design for “linear courtyard” Zhuangke under varying weights.

Performance - oriented	Weight of heating energy consumption	Building orientation (°C)	Courtyard length (m)	Courtyard width (m)	Building depth (m)	Building span (m)	Building net height (m)	South-facing window-to-wall ratio	Sunroom depth (m)	North double-layer wall cavity (mm)	Height above the ground(m)	Heating energy use intensity (kWh/m2)	Indoor standard effective temperature/℃
Thermal performance prioritised	0	15.0	16.7	8.5	5.9	15.5	3.0	0.47	2.0	400	1.0	15.35	12.11
	0.1	12.8	16.6	8.5	5.9	15.5	2.7	0.47	2.0	400	1.0	15.33	12.10
	0.2	−1.8	16.8	8.4	5.6	15.6	2.7	0.49	1.8	400	1.0	14.70	12.08
	0.3	13.0	16.7	8.3	5.9	15.5	2.7	0.50	2.0	400	1.0	14.42	12.07
	0.4	13.0	16.7	8.9	5.9	15.5	2.8	0.50	2.0	400	1.0	14.23	12.06
Target balance	0.5	13.0	14.1	11.3	5.7	13.5	2.4	0.54	1.8	400	1.0	13.89	12.05
Heating energy consumption prioritised	0.6	−2.7	16.5	11.3	5.7	15.3	2.5	0.54	1.8	400	1.2	13.56	12.03
	0.7	−2.7	16.5	11.3	5.7	15.3	2.4	0.54	1.8	400	1.2	13.50	12.02
	0.8	−2.7	19.4	11.4	5.9	18.6	2.7	0.48	2.1	400	1.2	13.15	12.00
	0.9	−2.7	19.1	11.4	5.9	18.3	2.7	0.49	2.1	400	1.2	13.11	11.99
	1.0	−2.7	19.5	11.4	6.3	18.6	2.5	0.51	2.1	400	1.2	12.72	11.96

As seen in Fig. 8, buildings facing south can receive the maximum amount of sunlight and increase solar radiation, thereby helping to enhance thermal energy collection in winter and reduce heating energy consumption. Longer courtyards can improve air circulation, which is conducive to reducing the demand for winter heating and thus influencing indoor temperature regulation. The courtyard’s width determines the area of sunlight exposure, and a more expansive courtyard may reduce heat loss on the building surface, thereby improving the thermal performance of the building and reducing heating demand. More profound buildings will affect the entry of natural light, reducing indoor lighting and increasing the burden of heating. Larger building spans may lead to difficulty achieving uniform indoor temperature distribution, thereby increasing the burden of heating. Sunrooms can provide additional heat by accumulating solar energy, reducing the demand for winter heating. The height of the building above the ground will affect the speed of heat loss in cold winds, resulting in higher heat loss and thus increasing the energy consumption for winter heating. A reasonable window-to-wall ratio helps maintain indoor warmth and reduce winter heating energy consumption. Northern double-layer wall cavities increase the thermal insulation performance of the walls and reduce the penetration of cold air. Deeper cavity walls can provide better insulation, slow heat loss, and reduce winter heating energy consumption.

As shown in Table 3, the spatial organisation is as follows: The building is oriented north-south, with an orientation angle ranging from − 15° to 15°. The length of the courtyard varies from 13.78 m to 18.75 m. The width of the courtyard ranges from 7.80 m to 10.10 m. Regarding the main spaces: The depth of the main building ranges from 5.03 m to 7.20 m. The bays of the main building range from 14.20 m to 16.70 m. The net height of the main building ranges from 2.60 m to 2.75 m. For the auxiliary space, the depth of the sunrooms ranges from 1.7 m to 2.5 m. Regarding the building interface, the window-to-wall ratio of the south wall ranges from 0.46 to 0.52. The cavity depth of the north double-layer fence ranges from 220 mm to 400 mm. The height of the base ranges from 0.75 m to 1.20 m.

Table 3.

Design parameter statistics for Zhuangke Dwellings.

Formal category		Design parameters	Range of values	Data statistical analysis chart
Space	Spatial organisation	Building orientation/°	−15~15
		Courtyard length (m)	13.78~18.75
		Courtyard width (m)	7.80~10.10
	Main space	Depth (m)	5.03~7.20
		Span (m)	14.20~16.70
		Height (m)	2.60~2.75
	Auxiliary space	Sunroom depth (m)	1.7~ 2.5
interface		window-to-wall ratio	0.46~0.52
		North double-layer wall cavity (mm)	220~ 400
		Height above the ground(m)	0.75~1.20

Based on the statistical results of the design parameters of the Zhuangke dwellings, the values of the design parameters for the benchmark model (Fig. 7) are respectively: the building orientation is 7.1°, the length of the courtyard is 16.2 m, the width of the courtyard is 8.8 m, the depth of the main building is 6.3 m, the span of the main building is 15.5 m, the net height of the main building is 2.7 m, the depth of the sunroom is 2.0 m, the window-to-wall ratio of the south wall is 0.50, the cavity depth of the north double-layer wall is 300 mm, and the height of the base is 0.75 m. Based on the simulation results, it can be known that the benchmark model’s initial unit floor area heating energy consumption for building performance is 15.50 kWh/m², and the indoor standard effective temperature is 11.49℃.

Multi-objective optimisation results

Based on the initial performance of the benchmark building, using the SPEA2 algorithm, the objective is to minimise the winter heating energy consumption and maximise the indoor thermal performance. The SPEA-2 algorithm can identify optimal trade-off solutions for diverse optimisation objectives and has been extensively applied in architectural multi-objective optimisation⁵². Therefore, SPEA-2 is selected to generate or approximate the Pareto-optimal solution set. Key parameters are configured as follows: elitism rate of 0.500, mutation probability of 0.048, mutation rate of 0.5, and crossover rate of 0.8. The population size is 40 to enhance computational efficiency for complex parametric models⁵³. The total number of cycles for optimisation) is 50. Finally, 25 Pareto solution sets for the one-way courtyard houses are obtained.

Figure 8 shows that all the optimal solutions on the Pareto frontier have lower winter heating energy consumption than the benchmark building. The winter heating energy consumption trend in is opposite to the total hours of thermal discomfort throughout the year. The two optimisation objectives are mutually restrictive and cannot be simultaneously optimised. This may indicate that the heating system is inefficient or that other environmental factors affect the indoor climate.

The TOPSIS comprehensive evaluation method is applied to evaluate two objective functions (the winter heating energy consumption and the weighted number of indoor standard effective temperatures) to obtain optimal solutions under different weights, as shown in Fig. 9. The sum of the weights of the two objective functions is 1. As shown in Fig. 10, the solid and dashed lines represent the corresponding building heating energy consumption and indoor standard practical temperature values of the optimal Pareto solution set under different weight schemes. It should be noted that the TOPSIS comprehensive evaluation method only acts on the optimal solutions at the Pareto front. When the weight is 0 or 1, the optimal solution of the objective function is not equivalent to the single-objective optimisation result.

Fig. 9.

Pareto set for “linear courtyard” Zhuangke.

Fig. 10.

Optimal solutions for “linear courtyard” Zhuangke under varying weights.

Figure 9 illustrates the Pareto front, revealing a clear conflict relationship between heating EUI and SET in the linear courtyard prototype: reducing heating energy consumption is inevitably accompanied by a slight decrease in comfort, while improving comfort requires higher energy input.

Figure 10 demonstrates how varying the weight assigned to energy conservation versus comfort alters the optimisation outcome. When energy efficiency weights increase from 0.0 to 1.0, SET decreases by 1.2% (from 12.11 °C to 11.96 °C), while EUI is reduced by 17% (from 15.35 to 12.72 kWh/m²). These findings confirm that passive design strategies in cold regions must be guided by goal-oriented trade-offs between energy efficiency and thermal comfort, allowing designers to adjust the balance flexibly based on specific performance priorities.

Table 4 presents detailed information on the optimal solutions for the “linear” Zhuangke dwelling under different weights of heating energy consumption. According to the variation range of the weight of heating energy consumption, the optimal solutions for courtyard houses can be classified into three categories: priority of thermal performance, multi-objective balance, and priority of heating energy consumption. The scheme with the best indoor thermal performance has an average value of the winter indoor standard effective temperature of 12.11℃, which is improved by 5.4% compared to the benchmark building. The winter heating energy consumption is 15.35 (kw ·h/m²), and the energy saving rate is 1%. The scheme with the most energy-saving energy consumption has a winter heating energy consumption of 12.72 (kw ·h/m²), which is 17.9% lower than that of the benchmark building. The average winter indoor standard effective temperature is 11.96℃, and the improvement rate is 4.1%. Under the balanced selection with a weight of 0.5, the winter heating energy consumption of this optimal solution is 13.89 (kw ·h/m²), the energy saving rate is 10.4%, the number of hours of thermal discomfort is 12.05℃, and the improvement rate is 4.9%.

Based on the data presented in Table 4, a qualitative assessment can be made regarding the influence of different design parameters on the building’s heating EUI and SET. Courtyard width exhibits the most significant impact on reducing EUI among all evaluated variables. An increase in courtyard width is associated with a clear downward trend in heating energy demand. Sunspace depth also shows a notable effect, influencing both EUI and SET. The south-facing window-to-wall ratio primarily enhances SET by increasing passive solar heat gains. Deviations in building orientation from true south slightly decrease SET and increase heating energy consumption, while variations in building net height and plinth height exert secondary effects on performance outcomes.

It should be noted that these conclusions are based on multi-parameter optimisation results, where multiple variables were adjusted simultaneously under different weighting scenarios. Therefore, at this stage, only qualitative trends regarding the influence of each parameter can be inferred, and it is not yet possible to accurately quantify the individual contributions of each variable to the performance outcomes. To achieve a more rigorous and quantitative evaluation of parameter impacts, future research should incorporate localised sensitivity analysis methods, such as the one-at-a-time (OAT) approach, or multivariate regression analysis. These methods would enable a more apparent distinction of the relative influence of each design parameter, thereby supporting more targeted passive design optimisation strategies.

Optimisation results analysis informs design strategy

Conduct an overall analysis of the optimisation results of the “linear” Zhuangke dwelling and summarise the design strategies from the perspectives of space and interface.

(1) Facing north and south can maximise solar energy utilisation and enhance buildings’ lighting and heating efficiency. The residential orientation range is from 15° east of south to 15° west of south.

(2) Enclosed courtyards significantly influence the solar radiation and natural ventilation conditions of buildings, affecting the heat exchange between indoor and outdoor spaces. According to the box plot and optimisation results, the recommended range for courtyard length is 13.78 to 18.75 m, with a preference for the median value within the surveyed data range of 8.8 to 22.1 m. For courtyard width, the recommended range is 7.80 to 10.10 m, with a preference for the median value within the surveyed data range of 6.1 to 13 m.

(3) Avoiding excessively high indoor clear heights is crucial, as it can lead to increased energy consumption and reduced thermal comfort. According to the box plot and optimisation results, the recommended range for main space height is between 2.60 m and 2.75 m, with a preference for the median value within the surveyed range of 2.3 to 3.0 m. This aligns with the provisions in the “Energy-saving Design Standard for Rural Residential Buildings, “⁵⁴ which stipulates that the indoor clear height of rural residences should not exceed 3 m.

(4) Excessive building depth must be avoided. A more considerable building depth leads to higher energy consumption. According to the box plot and the optimisation results, the building depth is distributed between 5.03 m and 7.20 m, with a peak around 6.0 m, which is the same as the recommended value of building depth in GB/T 50,824 − 2013 “Energy-saving Design Standard for Rural Residential Buildings.”

(5) Adding a sunroom can significantly enhance the absorption of solar radiation energy and function as an effective passive solar heating system during winter, thereby reducing heating energy consumption. According to the box plot and optimisation results, the recommended depth for the sunroom is between 1.7 and 2.5 m. The optimal value is towards the median of the surveyed data range (1.2 to 3.0 m), with an ideal depth of approximately 2.0 m.

(6) Moderately increasing the height of the platform base can effectively reduce heat loss from the building’s ground surface and enhance solar exposure and heat gain. However, an excessively high platform base may lead to increased construction costs. According to the box plot and optimisation results, the recommended range for the platform base height is 0.75 to 1.2 m, with a preference for values on the higher end of the surveyed data range (0.3 to 1.5 m).

(7) Increasing the south-facing window area helps fully utilise solar radiation heat in winter, improving the indoor average temperature and thermal comfort. According to the box plot and optimisation results, the recommended range for the south-facing window-to-wall ratio is from 0.46 to 0.52. It is suggested that the median value of the survey data be adopted (0.45 to 0.54).

(8) Increase the cavity space northward to reduce heat conduction and convection, enhancing the building’s thermal insulation performance. From the box plot and the optimisation results, the depth range of the north double-layer wall cavity is 220 to 400 mm.

Findings and discussion

Findings

This study collected the spatial and interface architectural design parameters data of rural Zhuangke dwellings in Qinghai Province, conducted statistical analysis, and optimised them with multiple objectives focusing on energy conservation and thermal performance. The research results indicated that the passive energy-saving design parameters of Zhuangke dwellings include courtyard length, courtyard width, building orientation, building height, building depth, building span, Height above the ground, window-to-wall ratio, sunroom depth, and the depth of the northern double-layer wall cavity. This study uniquely introduced a multi-objective algorithm, combining advanced spatial optimisation with parametric simulation tools, to optimise the passive energy-saving form design parameters of Zangke dwellings. Based on the statistical simulation optimisation results of the design parameters, strategies for guiding residential design in Qinghai Province were summarised.

This study contributes to the field by adopting a multi-objective, performance-oriented approach to optimise design parameters, bridging the gap between architectural form and its practical performance outcomes. The optimisation of the benchmark model led to significant improvements in energy conservation, with winter heating energy use intensity (EUI) reduced by 1–17.9% and enhancements in indoor thermal performance, with an average increase of 4.1–5.4% in the standard effective temperature. The study ensured accuracy and practical relevance in optimisation by integrating parametric modelling with on-site data. Furthermore, it focused on localised solutions by quantitatively extracting and optimising the passive energy-saving design principles of traditional dwellings in Huangyuan County, Qinghai Province, and applying them to residential design and renovation in the region. Unlike previous studies that generalised their findings to broader cold regions^3,36,55,56 this research highlights the potential of architectural design to optimise energy utilisation and thermal comfort, specifically in cold and resource-constrained regions, offering valuable reference forms and parameter values for residential architecture in Qinghai Province.

This study optimises the design parameters of traditional Zhuangke dwellings in Huangyuan County, focusing on energy saving and thermal performance, particularly in icy regions. The discussion will be based on different research regions and climates, design parameters and optimisation goals, and methods and tools to explore further the application of multi-objective optimisation in residential building design.

(1) Research area and climate adaptability.

While previous studies have explored cooling optimisation in tropical regions⁵⁷ energy-saving effects of courtyard spaces in rural China⁵⁸ and low-energy designs in temperate climates²⁰ this research addresses the challenges of icy regions. Unlike studies in hot-summer, cold-winter areas⁵⁶ which focus on balancing cooling and heating needs, this study emphasises reducing heating energy consumption and enhancing winter indoor thermal comfort, providing tailored strategies for harsh cold climates.

(2) Design parameters and optimisation objectives.

In multi-objective design, structural and environmental optimisation involves considering buildings’ structural and environmental performance to reduce energy consumption and improve comfort^59–61. Multi-objective optimisation in residential design often involves balancing structural, environmental, and economic factors. Studies have optimised prefabricated building components for cost, duration, and carbon emissions⁵⁴ or explored the impact of building shape, window-to-wall ratio, and roof slope on energy consumption²⁰. Multi-objective optimisation methods, such as artificial neural networks and metaheuristic algorithms, have also been used to balance building energy efficiency and indoor thermal comfort⁴. Similar optimisation strategies have been adopted to enhance buildings’ thermal performance and energy efficiency in cold regions⁷. While many studies focus on modern building designs or envelope modifications^5,62 this research uniquely emphasises optimising spatial and interface design parameters in traditional dwellings. Additionally, unlike studies that prioritise energy efficiency alone, this study integrates indoor thermal comfort as a critical objective, particularly for winter conditions in cold regions.

(3) Methods and optimisation tools.

Building performance simulation and multi-objective optimisation techniques (such as NSGA-II and PEA2 algorithms) have also been adopted^4,7,54,63. Some studies have integrated advanced algorithms like XGBoost to improve prediction accuracy⁹. Similarly, this study employs simulation tools and optimisation techniques to enhance energy efficiency and thermal comfort. However, it distinguishes itself by applying these methods to traditional dwellings, adapting design strategies to the specific climatic and cultural context of Huangyuan County.

In conclusion, while existing research provides valuable insights into multi-objective optimisation across various climates and building types, this study contributes a unique perspective by focusing on traditional Zhuangke dwellings in icy regions. It advances the field by optimising spatial and interface design with energy-saving and thermal comfort objectives. It offers a comprehensive approach tailored to the challenges of cold climates and traditional architectural forms.

Limitation

Although this study utilises a thorough approach, several limitations must be recognised.

In the first place, this study’s simulation was conducted based on the material properties of representative linear Zhuangke dwellings and the meteorological data of the typical meteorological year in the county. While TMY data is commonly used in building simulations, it represents an average climate year and may not capture a specific year’s variability or extremes in weather conditions. This can introduce uncertainty, as TMY data may not fully reflect actual temperature, humidity, or other environmental fluctuations, potentially affecting the robustness of the results. Moreover, the energy system and the situation of human activities were set according to the results of on-site interviews. These factors were obtained through limited data research and affected the accuracy of the simulation settings and results.

Additionally, due to the limitations of objective conditions, this study only focused on the Hehuang region, focusing on protecting traditional buildings in Huangyuan County. A total of 7 conventional villages in Huangyuan County and 53 courtyards were sampled for investigation. The database’s sample size was relatively small compared to many traditional dwellings in the Qinghai region. This limitation may affect the statistical reliability and generalizability of the identified optimal design parameters, as the limited sample size could lead to the omission of edge-case architectural features and parameter variations, thus failing to fully capture the diversity of design characteristics across the region. Consequently, unavoidable discrepancies may arise in the derivation and statistical analysis, such as reduced statistical significance of outliers and wider confidence intervals for the design parameters. Future work should include a larger number of samples to address this limitation.

Moreover, the residential houses simulated in the study were the most typical linear courtyard-type Zhuangke dwellings. Other common typologies were not included, such as L-shaped and U-shaped layouts. These alternative forms differ significantly in spatial organization, degree of enclosure, and internal heat distribution, influencing key design parameters such as window-to-wall ratio, courtyard geometry, and building volume. Such morphological differences may result in distinct energy consumption patterns and thermal performance, as the mechanisms of passive heat gain and loss respond differently to similar climatic conditions. These variations merit further in-depth investigation.

While SET is suitable for evaluating non-steady-state thermal environments, it has certain limitations under cold, low-airflow conditions. Future research will benefit from comparative analysis with operative temperature and adaptive models, along with the integration of detailed occupant behaviour patterns (e.g., window operation, clothing adjustment), to enhance the robustness and contextual relevance of thermal comfort evaluations.

Ultimately, this study conducted quantitative optimisation on the dwelling’s design parameters, with the building’s material structure and equipment set according to the actual conditions of traditional houses, without incorporating advanced technologies such as new materials and energy systems. Combining traditional passive design strategies with modern technologies to enhance buildings’ energy efficiency and thermal performance is also of significant importance.

Conclusion

This study optimised the design parameters of the Zhuangke dwellings in the Huangyuan County of Qinghai Province under cold, semi-arid climate conditions, focusing on energy conservation and thermal performance. Ten design parameters were extracted and summarised through on-site investigations, green building norms and residential prototype theory. Box plots were used for statistical analysis on 53 data sets, determining each parameter’s value range. A benchmark model of a typical linear quadrangle courtyard residence was constructed using a parametric platform, with the parameter values set at the median of their respective ranges. The architectural construction parameters and operating conditions were set based on the investigations and norms. Then, the design parameters of the “linear courtyard” Zhuangke dwellings were optimised for multi-objective performance (energy conservation and thermal performance) using the parametric platform.

The research results indicate:

(1) Under the conditions of cold winter and intense solar radiation, the design parameters of the Zhuangke dwellings in Huangyuan County exhibit a climate response mechanism for reducing heat loss and increasing heat gain. These design parameters include:

(2) A box plot statistical analysis was conducted on the design parameter values of 53 rural dwellings. Extreme values were removed, and the design parameter ranges of each prototype were determined by referring to the maximum value, minimum value, and data dispersion—orientation: 15° south of east to 15° south of west. Courtyard dimensions include a depth ranging from 7.8 m to 10.10 m and a span from 13.78 m to 18.75 m. The main building spans 14.20 m to 16.70 m, a depth of 5.03 m to 7.20 m, and a height of 2.60 m to 2.75 m. The depth of the double-layer cavity in the north wall ranges from 220 mm to 400 mm. The sunroom depth varies between 1.7 m and 2.5 m. The window-to-wall ratio for the south-facing wall is set at 0.46 to 0.52. The platform height above ground level is between 0.75 m and 1.2 m.

(3) The simulation optimisation results show that the optimised design parameters of the benchmark model have reduced the winter heating energy consumption to varying degrees and improved the indoor thermal performance in winter. The influence degree of each design prototype on the SET of the building’s indoor thermal performance ranges from 0.47 to 0.62℃, and the influence degree on the EUI of the building ranges from 0.15 to 2.78 kWh/m².

Considering the findings and limitations of this study, several potential directions for future research can be identified.

Future research could broaden the scope by including regions with varying climatic conditions beyond Zhuangke dwellings in Qinghai Province to enhance the diversity of samples. By analysing diverse traditional dwellings, the universality of the multi-objective optimisation method proposed in this study can be validated, and region-specific design strategies can be developed. For instance, comparative studies could be conducted in other areas of the Qinghai-Tibet Plateau, arid regions in Northwest China, or humid regions in Southern China to explore the variations in optimal design parameters under different climatic conditions.

While this study primarily focused on energy efficiency and thermal performance, future research could integrate additional optimisation factors, such as economic costs, material sustainability, cultural suitability, and residents’ behavioural patterns. A holistic optimisation approach would provide more comprehensive solutions for practical engineering applications. For example, multi-objective optimisation algorithms (e.g., NSGA-II) combined with life cycle assessment (LCA) methods could be employed to quantify the impacts of different design parameters on economic, environmental, and social aspects.

To deepen the research on external climate and dynamic environmental conditions in simulation studies, in the future, annual dynamic meteorological data can be used instead of the data of the coldest week to evaluate the dynamic changes of yearly energy consumption and thermal performance of buildings.

To combine material innovation and renewable energy technologies, modern building materials and renewable energy technologies (such as solar photovoltaic systems) are combined with traditional spatial forms to enhance the potential for energy conservation.

Advancements in simulation tools and multi-objective algorithms, mainly through integrating artificial intelligence or machine learning, can enhance the precision and scalability of building performance evaluation and design parameter optimisation.

Overall, this study highlights the crucial role of design parameters in the energy efficiency and thermal performance management of Zhuangke residences in Huangyuan County, Qinghai Province, under cold climate conditions. It has emphasised the significance of multi-objective optimisation methods in enhancing the passive energy-saving design of Zhuangke dwellings. Through extensive field investigations and statistical analyses, this study has determined specific value ranges for the design parameters of “linear courtyard” dwellings and summarised design strategies. It has practical significance for improving the energy-saving design of linear courtyard-type Zhuangke dwellings in rural areas of Qinghai Province and similar cold climate regions.

Abbreviations

ANN

Artificial neural network

CSMD

China standard meteorological database

EUI

Energy use intensity

MDO

Multidisciplinary optimisation

MOPSO

Multi-objective particle swarm optimisation

OAT

One-at-a-time (sensitivity analysis method)

PMV

Predicted mean vote

RER

Renewable energy ratio

RH

Relative humidity

SET

Standard effective temperature

SPEA2

Strength pareto evolutionary algorithm 2

TMY

Typical meteorological year

TOPSIS

Technique for order preference by similarity to ideal solution

Author contributions

Y.L.: Conceptualisation, Formal analysis, Investigation, Methodology, Validation, Visualisation, Data curation, Writing an original draft. Y.C.: Review & editing. L.Y.: Software, Investigation,, Review & editing. D.H.: Review & editing. M.A.: Supervision.

Data availability

The datasets used and analysed during the current study are available from the corresponding author upon reasonable request.

Declarations

Competing interests

The authors declare no competing interests.