An AI-driven multi-objective framework for optimizing window dimensions considering energy demand and thermal comfort

Why Window Size Matters More Than You Think

When we picture energy-efficient buildings, we often imagine solar panels or thick insulation—but ordinary windows quietly make or break how much energy a building uses and how comfortable it feels. This study looks at a deceptively simple question with big real-world stakes: how wide and tall should an office window be so that people inside feel comfortable while the building’s heating and cooling bills stay in check? By using artificial intelligence to sift through thousands of design options, the researchers show how “just-right” windows can strike a smart balance between comfort and energy use in a hot, dry city like Tehran.

A Simple Building to Answer a Tricky Question

To untangle the effects of window size, the authors start with a deliberately simple one-story office room shaped like a shoebox. The walls, roof, floor, and building orientation are all fixed, and there is only one window on the main façade. The only things that change are the window’s width and height, which are varied over a realistic range from quite small to very large. For each of nearly ten thousand window combinations, computer simulations estimate how much energy the building needs for cooling and heating over a year, and how often indoor temperatures fall within a comfort band that people are likely to find acceptable without air conditioning. This stripped-down setup lets the team focus squarely on how window size alone reshapes energy demand and comfort.

Letting an Artificial Brain Learn the Patterns

Running thousands of detailed simulations is time-consuming, and exploring even more design options quickly becomes impractical. To speed things up, the researchers train an artificial neural network—a type of AI loosely inspired by how brains process information—to learn from the simulation results. Once trained, this “surrogate” model can instantly predict cooling and heating needs and comfort levels for any new window size within the studied range. Tests show that its predictions track the original simulation results extremely closely, capturing more than 99% of the variation in all three measures. In other words, the AI becomes a fast, trustworthy stand-in for the slower physics-based simulations.

Searching for the Best Trade-Offs, Not One Perfect Answer

With this fast digital stand-in in place, the team turns to evolutionary search methods—algorithms inspired by natural selection—to hunt for window sizes that balance competing goals. Here the goals are to reduce yearly cooling energy, reduce yearly heating energy, and increase the share of time when indoor temperatures feel comfortable without mechanical cooling. Because making windows bigger helps in winter (more sun and warmth) but hurts in summer (more overheating), there is no single “best” window. Instead, the algorithms build a family of equally good compromise designs, known as a Pareto front, where no option can be improved in one goal without making another worse. For this particular office in Tehran, windows of intermediate size emerge as the sweet spot: they keep heating demand much lower than small windows, keep cooling demand more reasonable than the largest windows, and provide comfort levels above roughly 80% of occupied hours.

What the Results Reveal About Window Size

The patterns that emerge from the AI-driven search are intuitive yet quantifiable. As window area grows, annual cooling energy climbs almost steadily because the south-facing glass admits more solar heat during the hot season. At the same time, heating energy drops because winter sunlight provides useful warmth that more than compensates for heat loss through the glass. The comfort measure used in this study—how often indoor temperatures fall within an adaptive comfort band for naturally ventilated spaces—also rises with window size, mainly because larger windows capture more mild-weather warmth. However, this index does not fully penalize brief or extreme overheating, so very large windows might still feel uncomfortably hot at times, even if they score well on the chosen comfort metric. This nuance underlines why designers should treat the study’s comfort results as comparative guidance rather than as absolute guarantees.

How This Helps Designers and Building Owners

For architects, engineers, and even building owners, the practical message is straightforward: window sizing is not a matter of “bigger is always better” or “small is always safer.” Instead, the study shows that there is a band of intermediate window dimensions that delivers strong comfort benefits while keeping energy demand in check, especially in climates with long heating seasons and shorter, intense cooling periods like Tehran. By combining detailed simulations, a fast-learning AI model, and evolutionary search, the framework gives designers a clear menu of high-performing options rather than a single rigid prescription. That means they can weigh comfort, energy bills, and aesthetic preferences side by side—using data, not guesswork—when deciding how big to make the windows in future buildings.

Citation: Nasab, S.M.R.A., Rabiei, H. An AI-driven multi-objective framework for optimizing window dimensions considering energy demand and thermal comfort. Sci Rep 16, 6365 (2026). https://doi.org/10.1038/s41598-026-39795-8

Keywords: window design, building energy, thermal comfort, artificial intelligence, multi-objective optimization