Units
1. Introduction to Artificial Intelligence in Architecture: Boundaries, Validation, Responsibility and Ethics 2. Visual Production with Concept Design and Artificial Intelligence 3. Area Programme, Layout and Plan Optimization 4. BIM Integration: Workflow with Revit, IFC and Artificial Intelligence 5. Energy, Daylight and Comfort Analysis 6. Regulation, Zoning and Legislation Research and Verification 7. Rendering, Visualization and Presentation 8. Quantity Surveying, Cost Estimation and Discovery 9. Sustainability, Material Selection and Life Cycle 10. Structural and Security-Critical Decisions: Approval and Limits 11. Construction Site, Project Management and Documentation 12. End-to-End Artificial Intelligence Supported Architectural Workflow and Office Application
Unit 5 / 12

Energy, Daylight and Comfort Analysis

Gains:

  • Ability to describe the role of AI in data preparation, interpretation and reporting in the daylight, insolation and energy simulation workflow
  • Ability to clearly track assumptions, boundary conditions, and uncertainty when summarizing simulation results with AI
  • Ability to understand that energy/comfort results must be verified with official calculation method and accredited software and that AI does not produce numbers

A building doesn't just look beautiful; It keeps the person living inside cool in the heat and warm in the cold, lets in light during the day, and does not consume unnecessary energy. To anticipate these, architects use simulations: computer analyzes that numerically calculate how the building will behave with the sun, light and air. Daylight analysis (how much natural light enters the space), insolation/shade analysis and energy simulation are the main ones. Accredited and verified special software makes these calculations; Artificial intelligence does not produce these numbers itself. The role of AI is to prepare data, organize scenarios, interpret and report results. In this unit you will learn how to put AI in the right place in the energy and daylight workflow and how to avoid the most dangerous mistake: making AI produce numbers.

Why AI doesn't simulate

A language model can give a confident number to the question "how much sunlight gets into this room". But this number is an estimate/fabrication, not a real calculation. Daylight multiplier, energy load, shadow hours; These are physical calculations based on the building's geometry, geographical location, material properties and climate data. Only simulation software that models physics gets these things right.

The real contribution of AI is this: preparing the simulation's input (scenario list, bill of materials), making sense of its output (summarizing hundreds of pages of results, finding patterns), and writing its report. So AI is valuable before and after the account; not in the account itself.

Caution: Do not ask the AI ​​for a number such as daylight multiplier, energy consumption, or shadow hour. Only accredited/verified simulation software produces these values. Any number provided by AI should not be used unless it is a sourced account.

Preparing input: The first benefit of AI

The quality of the simulation depends on the quality of the input. AI helps you orchestrate a complex set of scenarios: which façade directions, which glazing types, which shading options, which usage hours to test. Turning these into a structured table and clarifying what each scenario will measure saves time and reduces errors.

AI can also compile material properties (heat conductivity value, light transmittance, etc.) and put them in a table; but it is necessary to confirm these values ​​​​from the manufacturer's documentation, because the AI ​​may remember incorrectly.

Interpreting output: The second benefit of AI

The simulation output is often huge and complex: hour by hour temperature, room by room light level, monthly energy distribution. Given this data, AI produces a clear summary: "the three places that get warmest", "areas below the daylight target", "in which month the cooling load is highest". This brief expedites the architect's design decision. But be sure to specify which software and with what assumptions you are giving the results to the AI; The comment should be placed in this context.

Tip: When giving a simulation result to the AI, say "these results were produced by [the software], with the following assumptions; you just comment and indicate how the assumptions affected the result." AI thus makes uncertainty visible rather than hiding it.

Step by step: AI-powered analytics workflow

  1. Clarify the question. What do you want to know: overheating, lack of light, energy?
  2. Edit scenarios with AI. Table the variants to be tested.
  3. Compile and verify material data. They say AI, you verify it from the manufacturer's documentation.
  4. Run the simulation with competent software. This is where the number comes from.
  5. Interpret the output with AI. Make a summary by stating the assumptions.
  6. Make the decision and report. The conclusion and justification are recorded with its source.

three mini cases

Case 1 — Scenario editing. An architect wants to test daylight for 4 facade orientations, 3 glazing types and 2 shading options. AI turns these combinations into a clear table of 24 scenarios, and the architect sees which ones are unnecessary repetition and narrows it down to 16. Simulation time and irregularity are reduced.

Case 2 — Interpretation and pattern. 800 lines of hourly temperature output appears. The architect gives this to AI and asks for a summary. AI marks two offices in the south-west corner warming above target in the afternoon. To verify this finding, the architect looks at the corresponding graph of the simulation and adds shading. AI did not produce the number here; It made the existing number readable.

Case 3 — Dangerous shortcut. To save time, an intern asks the AI ​​"what is the daylight multiplier of this room?" and puts the resulting number in the report. Upon checking, it becomes clear that this number is not based on any calculations and is made up. The report is withdrawn. Lesson: numbers always come from competent software.

Four copyable prompts

Turn the SCENARIOS to be tested for the following analysis target into a table. Columns: scenario no | front direction | glass type | shading |what does it measure? Flag unnecessary/repetitive scenarios. Aim: [...]

Compile the heat transmittance and light transmittance values ​​for this material list in a table. Mark each value as "TO BE CONFIRMED FROM MANUFACTURER'S DOCUMENT"; Don't give exact value. List: [...]

The following simulation results were produced by [software name], with the following assumptions: [assumptions]. SUMMARY results: 3 most critical places, patterns and impact of assumptions. DO NOT GENERATE new numbers, just interpret the given ones. Results: [...]

Organize this energy/daylight report draft in an architect's language: with sections for method, assumptions, findings, recommendations. Leave a [SOURCE] field to indicate the source (software/account) of each issue.Draft: [...]

Weak prompt / Strong prompt

Weak: "What would be the energy consumption of this room?"

Strong: "I'm pasting the monthly energy output of my EnergyPlus-based simulation below. Assumptions: [climate file, usage hours, set temperatures]. Summarize these results: which month has the highest load, which location stands out, which of the assumptions affects the result the most. Calculating a new number; just interpret the data I gave."

Powerful prompt keeps AI in interpreter role; It does not fall into the trap of producing numbers and makes assumptions visible.

business

Does AI do it?

Who produces the number

Script editing

Yes

—

Material data compilation

Yes (confirmed)

Manufacturer's certificate

simulation account

no

Accredited software

Concluding comment/summary

Yes

—

Report writing

Yes (welded)

—

Common mistakes

  • Making AI produce numbers. Daylight/energy value comes from software only.
  • Not stating assumptions. The interpretation remains without context and misleads.
  • Not confirming material value. AI may give incorrect temperature/light values.
  • Not comparing the AI ​​summary to the raw data. Summary may highlight incorrectly.
  • Putting the made-up number on the report. No unsourced numbers are included in the report.

In summary

Energy and daylight calculations are based on physics and only accredited simulation software gets it right. AI doesn't do this calculation; edits the input, compiles (verifies) the material data, interprets the result, and writes the report. Never make the AI ​​generate a number like daylight multiplier or energy consumption. Cite the source of each number, make assumptions visible, and verify the final assessment with competent expert.

Application task

Design a daylight or overheating analysis scenario for a space. With AI, turn test cases into a table and leave confirmation fields for material values. If you have a simulation output (or a sample table), summarize it to the AI, stating the assumptions, and check that the AI ​​does not produce any new numbers.

checklist

  • [ ] I clarified the analysis question.
  • [ ] I edited the scenarios with AI.
  • [ ] I confirmed the material values ​​from the manufacturer's documentation.
  • [ ] I got the number only from accredited software.
  • [ ] I had the AI ​​interpret it by stating the assumptions.
  • [ ] I checked that the AI ​​is not generating new numbers.
  • [ ] I wrote the source of each issue in the report.
Previous Next