Optimising building efficiency with genetic algorithms

Introduction

The use of digital twins in the planning, construction and operation of buildings was outlined in the previous post. The following discussion touches on the use of genetic algorithms and related software to optimise building design and performance. A selection of papers describing the use of genetic algorithms in this way will be cited together with references to background information.

Genetic Algorithms

The origins of software based on natural selection have been traced back to early computer scientists such as Alan Turing, John Von Neumann and Norbert Weiner (Mitchell, 1995). Genetic Algorithms (GAs) have found application in many areas where more conventional methods are difficult to apply, such as problems involving many variables. The ‘solutions’ proposed by a GA need to be evaluated, and this may require additional software specific to the problem in hand, in this case the computer modelling of buildings.

Building Information Modelling

The history of building information modelling (BIM) is outlined by BIM & SCAN (2025). Computer-aided design (CAD) systems were available in the 1970s: “By the 1990s and early 2000s, the concept of BIM gained traction as computing power increased and software developers began creating more advanced modelling tools.” Recognition by industry and governments led to BIM adoption in many countries “driving its integration into standard construction practices.” EnergyPlus software dates from c. 1999 and became open source in 2012. It is “a whole building energy simulation program that engineers, architects, and researchers use to model both energy consumption—for heating, cooling, ventilation, lighting and plug and process loads—and water use in buildings” (EnergyPlus, 2025). All the papers on optimisation cited below refer to this software.

Building optimisation research

 An early building performance problem concerning the effects of sunshine entering a building is described by Evins (2010). Energy from solar radiation trapped by the building is known as solar thermal gain. In summer too much solar gain “causes overheating and increases the need for cooling; too little solar gain in winter increases the need for heating.” The amount of artificial lighting needed also depends on the amount of sunshine and the position of the building. The problem was approached using EnergyPlus and a genetic algorithm (GA). To show that the GA was effective, its results had to be compared with a trusted method. A single room with defined dimensions in a stated geographic location was modelled, together with its windows, lighting, orientation, interior temperature control settings, equipment use, and number of occupants. Each of the variables could be altered in defined steps, giving several thousand possible combinations. EnergyPlus was then used to calculate the energy that would be used over a twelve-month period, using weather information for the chosen location, for every combination of the variables. This was described as a “brute force” method of finding the combination that gave the optimum solution in terms of cost and energy use. The genetic algorithm could then be given the same problem, using EnergyPlus to test each of its proposed solutions, with the intent of reaching the optimum solution much more efficiently than by the brute force method. The conclusion of the tests showed that the GA was potentially a useful tool in problems of this kind.

A longer and more detailed paper from the following year mentions the “growing need for building elements analysis and searching for optimal design using energy performance simulation” in the design stages. (Oh, S-M, et al., 2011). Again, a genetic algorithm is used in conjunction with EnergyPlus, but the problem is more complex and the method more refined than that described by Evins.  The study modelled a five-storey library building with a total floor area exceeding 7000 m². Two performance criteria were chosen, the energy use and an index of thermal comfort. The index used was the Predicted Mean Vote (PMV), which attempts to predict the average thermal sensation of the occupants. In problems with two or more performance criteria, the GA may produce multiple solutions which have to be subjected to further selection procedures. As a first stage the authors used Pareto optimality, where solutions were rejected if an improvement in one performance criterion resulted in worsening in another. The smaller resulting set of solutions can be further reduced by the subjective judgments of a human decision maker (DM).

A 2017 research paper claimed that in many countries the building sector accounted for about 40% of the total energy consumption, and that early design decisions “have a significant impact on the energy performance of buildings.”  The paper described the multi-variable optimization of selected design parameters in a single-family building which required both heating and cooling due to its geographical location. The parameters included type and size of windows, building orientation, insulation of external walls, roof and ground floor, and infiltration. Their effects on the life cycle costs (LCC) were analysed using genetic algorithms and EnergyPlus supported by MATLAB (Ferdyn-Grygierek, 2017). The simulation program modelled the heated and unheated rooms, internal heat gains from the occupants, their computers, TV set, kitchen equipment and lighting. An hourly schedule for occupation and lighting was used for each room. Solar radiation and weather data were provided for the location (Katowice), and the simulation ran with a fifteen-minute time step. Varying the model showed significant changes in life cycle costs and demonstrated the importance of choosing the best position for windows; the need to take account of properties other than their heat transfer coefficients; the dependence of optimum glazing on the external wall insulation; and the need to make decisions on cooling systems prior to determining optimum qualities of a building.

Lin and Yang (2018) used building simulation to compare the performance of designs made to current energy efficiency standards with their performance after optimisation using a multi-objective genetic algorithm. Five city locations were chosen in the hot summer and cold winter region of China. “The trade-offs between the annual energy consumption (AEC) and initial construction cost, as well as between life cycle cost (LCC) and number of thermal discomfort hours” were explored using a Non-dominated Sorting Genetic Algorithm (NSGA-II). These algorithms use sorting techniques “crucial for identifying and preserving high-quality solutions in multi-objective optimization” (GfG, 2025). The reference building is a “single-story concrete frame residential building”, and the study is mainly concerned with the “building envelope and the cooling and heating setpoint” and does not seek to optimise the Heating, Ventilation, and Air Conditioning system; it assumes a constant wall-to-window ratio, and a limited range of other building parameters. Within these limitations, some of the main results are as follows. The average energy saving potential compared to the building designed using the current energy efficiency design standard was nearly 30% and could exceed 38% with an increase in building cost of just over 3%. The key parameters affecting the building’s energy consumption are wall-to-window ratio, the heating and cooling temperature setpoints, roof insulation and external wall insulation. The optimization approach was considered suitable both in the building design phase and in planning retrofit for existing buildings.

Alexakis et al. (2025) describe work in progress on software development aimed primarily at building retrofit projects. They present “the conceptual design of an integrated simulation pipeline that couples Genetic Algorithms (GAs) with dynamic building performance simulations.” The pipeline “utilises microclimate-specific weather data rather than standardised datasets” and “objective functions are … based on stakeholder inputs, allowing the tool to reflect diverse priorities, such as energy savings, economic feasibility, or comfort improvements.” EnergyPlus will provide the energy and comfort simulations, and the NSGA-II and NSGA-III genetic algorithms will be used for multi-objective optimisation. After simulation and optimisation, scenarios will be “ranked through a Multi-Criteria Decision Support System (MDSS), providing tailored recommendations.” This system can be configured by users to assign their weights or preferences to the various objectives. “Post optimisation, the MDSS ranks the Pareto-optimal solutions accordingly, presenting the five highest-ranked scenarios with their associated performance indicators.”

Alaili et al., 2025, address the problem of “designing buildings that enhance human comfort while minimizing energy consumption” from both heating and cooling demand. Their three-step approach simulates energy use by means of EnergyPlus software, produces sets of solutions using the NSGA II genetic algorithm, and employs multi-criteria decision analysis (MCDA) to rank these solutions. The base model was a single-story family house, and software simulations were used for three different climatic regions in France. The decision variables in each case included “wall insulation material properties, building rotation, glazing type, window-to-wall ratio (WWR), heating and cooling set-point temperatures, and infiltration rate.” The optimisation process yielded different numbers of Pareto solutions for each climate zone, and then additional considerations were imposed such as building standards, Passivhaus standards and client preferences. Decision analysis used the method of ranking with multiple reference profiles (RMP), and the energy use showed reductions on that of the base model in the range 6% to 22%. The authors claim that their ranking method has not previously been applied to optimisation problems of this kind.

A discussion of modelling real world decision problems can be found in Olteanu et al., 2019, who contrast ranking with choice and sorting. Background to decision making in multicriteria problems is provided in Roy, 1996. 

References

 Alaili, K., et al., 2025, Optimizing building envelope design across various French climates: A multi-objective approach using NSGA II and RMP method, Springer, abstract online, accessed 17 February 2026

https://link.springer.com/article/10.1007/s12273-025-1280-4

Alexakis, K., et al., 2025, Towards Optimal Building Retrofits: A Multi-Objective Optimisation Framework Using Genetic Algorithms and EnergyPlus, CEST 2025, online, accessed 18 February 2026

https://cms.gnest.org/sites/default/files/Proceedings/cest2025_00309/10377/cest2025_00309.pdf

BIM & SCAN, 2025, The History of Building Information Modelling (BIM), online, accessed 18 February 2026

https://bimandscan.com/history-building-information-modelling-bim/

Evins, R., 2010, Configuration of a genetic algorithm for multi-objective optimisation of solar gain to buildings, Buro Happold and University of Bristol, online, accessed 14 February 2026

https://www.bristol.ac.uk/media-library/sites/eng-systems-centre/migrated/documents/ralph-evins.pdf

EnergyPlus, 2025, EnergyPlus, online, accessed 18 February 2026

https://energyplus.net/

Ferdyn-Grygierek, J. and Ferdyn-Grygierek, K., 2017, Multi-Variable Optimization of Building Thermal Design Using Genetic Algorithms, 2017, Energies, online, accessed 14 February 2026

https://www.mdpi.com/1996-1073/10/10/1570

GfG, 2025, Non-Dominated Sorting Genetic Algorithm 2 (NSGA-II), GeeksforGeeks, online, accessed 14 February 2026

https://www.geeksforgeeks.org/deep-learning/non-dominated-sorting-genetic-algorithm-2-nsga-ii/

Lin, Y., and Yang, W., 2018, Application of Multi-Objective Genetic Algorithm Based Simulation for Cost-Effective Building Energy Efficiency Design and Thermal Comfort Improvement, Frontiers in Energy Research, online, accessed 14 February 2026

https://www.frontiersin.org/journals/energy-research/articles/10.3389/fenrg.2018.00025/full

Mitchell, M., 1995, Genetic Algorithms: An Overview, John Wiley & Sons, online, accessed 19 February 2026

https://onlinelibrary.wiley.com/doi/pdf/10.1002/cplx.6130010108

Oh, S-M, et al., 2011, PROCESS-DRIVEN BIM-BASED OPTIMAL DESIGN USING INTEGRATION OF ENERGYPLUS, GENETIC ALGORITHM, AND PARETO OPTIMALITY, Proceedings of Building Simulation, online, accessed 14 February 2026

https://publications.ibpsa.org/proceedings/bs/2011/papers/bs2011_1354.pdf

Olteanu et al., 2019, Preference Elicitation for a Ranking Method based on Multiple Reference Profiles, HAL Open Science, online, accessed 17 February 2026

https://hal.science/hal-01862334/file/Preference_Elicitation_for_a_Ranking_Method_________based_on_Multiple_Reference_Profiles.pdf

Roy, B., 1996, Multicriteria methodology for decision aiding, Kluwer Academic, online, accessed 17 February 2026

https://books.google.co.uk/books?hl=en&lr=&id=hT23RXsS8bQC&oi=fnd&pg=PR13&dq=B.+Roy.+Multicriteria+Methodology+for+Decision+Aiding.

Comments

Post a Comment

Popular posts from this blog

Directed technology and the environment

The subject of the previous post was Daniel Susskind’s book Growth : A Reckoning . Its author outlined measures that he believed could enable continued economic growth that did far less harm to society and the environment than has previously been the case. An important factor in this kind of growth would be steering technological progress in desirable directions through measures such as economic incentives. Susskind treats this idea broadly, but a more detailed exploration is found in the paper Environmental Policies and directed technological change by Gugler, Szücs and Wiedenhofer (2024). The authors examine how different environmental policies “can steer innovation towards eco-friendly technologies”, and produce numerical estimates of their effectiveness. The policies evaluated are those “designed to address environmental market failures”. Economists have seen climate change as a type of free market failure: a failure to maximise society’s welfare, since greenhouse gas emissions ar...
Read more

Growth and its Tradeoffs

Daniel Susskind’s 2024 book Growth: A Reckoning has received some enthusiastic reviews. Kate Barker described it as “an excellent book, developing a clear argument and not afraid to look really big questions squarely in the eye” (Barker 2024).  Susskind is a professor at King’s College, London, and Robert Bellafiore Jn. wrote that his book “considers the debates over growth’s causes, its recent elevation as a priority for governments, policies for promoting it, and rising concerns about its downsides” describing the work as “a concise and informative study of the idea, its past, and its potential future” (Bellafiore, 2024).  Alexander Bishop wrote at greater length in his article “Why not degrowth?” He noted that for Susskind growth “is a recent development. For most of the world’s history, from when humans started farming until roughly 1800, there was no meaningful growth. This is the period that Susskind calls “the great stagnation”. It is only since the industrial revoluti...
Read more

Biodiversity and clean energy

Introduction A speaker in a recent discussion on environmental issues remarked that biodiversity always has priority over clean energy. Whatever the truth of this statement, it prompts the question of how we should try to resolve situations where clean energy and biodiversity are in conflict. Certainly the frequent mention of biodiversity in the Sustainable Development Goals of the United Nations supports its importance, but clean energy also has a role in efforts to control climate change, and compromise may be the best policy if clean energy projects conflict with efforts to preserve and promote biodiversity. Such conflicts and their possible resolution will be examined below (CBD, 2018; Global Goals, 2021). Biodiversity and its importance Biodiversity broadly means “all the different kinds of life” in a given area, “the variety of animals, plants, fungi, and even microorganisms like bacteria that make up our natural world”. It is important to us because it “supports everything...
Read more