EES optimization methods are sufficient when you have one objective, bounded variables, and a license with enough optimization capacity. Choose parallel multi-objective search when you need a Pareto front, repeated EES solves, or more throughput than serial Genetic or DIRECT runs can provide.
The practical decision depends on three limits: your EES license, your number of design variables, and the number of competing objectives. F-Chart documents the built-in limits. Multi Optimization documents a separate workflow that controls EES.exe, runs private per-core folders, and evaluates up to 99 objectives with up to 20 design variables.
Which EES optimization methods fit your model?
Use built-in EES optimization when one objective and a serial solve budget match your model. The current F-Chart comparison shows a sharp difference between Student, Commercial, and Professional capacity.
| License | Price | Equation limit | Multi-variable methods | Optimization-variable limit | Use when |
|---|---|---|---|---|---|
| Student | $160 | 3,000 | 0 | 0 | Teaching, equation testing, and single-variable studies |
| Commercial | $600 | 6,000 | 2 | 20 | Small engineering models with limited multi-variable search |
| Professional | $1,200 | 12,000 | 5 | 80 | Larger models and broader built-in optimization studies |
That EES license comparison is a decision boundary, not just a price table. Student has no listed multi-variable methods or optimization variables. Commercial supports two methods and 20 variables. Professional raises those limits to five methods and 80 variables, but the workflow remains centered on EES’s built-in optimization process.
If the model also feeds a larger system study, keep the surrounding simulation separate from the EES choice. For example, use trnsys software for building-system coupling, then reserve EES for the component equations that need thermodynamic optimization.
What does EES Min/Max optimization actually control?
EES Min/Max optimization controls a single objective over a bounded design space. The EES Manual requires every independent variable to have finite lower and upper bounds, and the run stops when it reaches the relative tolerance or the maximum function-call count.
What should you check in your EES optimization variables?
- Give every independent variable finite lower and upper bounds before starting the search.
- Set a function-call limit that fits the cost of property calls and equation convergence.
- Use relative tolerance to stop when additional improvement becomes smaller than the engineering decision requires.
- Record failed evaluations separately from valid objective values so penalties do not look like physical results.
The Manual describes Genetic as robust but slow. It also describes DIRECT as a method that can locate global optima more efficiently in suitable bounded spaces. Both approaches are still serial from the standpoint of a single EES optimization run. If you need several objectives, you must either scalarize them or repeat the workflow.
That distinction matters in practice. A weighted objective can hide a useful design that performs well on one metric and poorly on another. Read multi objective optimization when you need the Pareto-front decision logic before choosing weights.
When is parallel multi-objective EES the better choice?
Choose parallel multi-objective EES when you need several competing objectives evaluated from the same model. Multi Optimization documents `$Import` and `$Export` data exchange, `EES.exe /solve` control, and private work folders for each core.
The documented range is 1–99 objectives and up to 20 design variables per model. That is different from the built-in EES license limits because the external workflow distributes independent model evaluations instead of asking one EES session to perform the entire search serially.
Use the same pattern for coupled studies only when the coupling is explicit. TRNSYS Energy System Simulation Explained is the next reference if your EES model sits inside a time-dependent energy-system calculation. For throughput planning, How to Run TRNSYS Simulations in Parallel (and Cut Optimization Time by ~8×) shows the parallel execution issue from the system-simulation side.
What did the real EES benchmark measure?
Our benchmark favored the parallel workflow because the model had two objectives and 240 independent evaluations. We used the real `R245fa_ORC_case.ees` thermodynamic model with three design variables: evaporator pressure from 1.20 to 2.40 MPa, condenser pressure from 0.15 to 0.35 MPa, and recuperator effectiveness from 0.60 to 0.90.
The objectives were to maximize net power and minimize total heat input. Each run used a population of 48 for five generations, giving 240 solves. Eleven runs failed. The primary failure trigger was an invalid logarithm when the modeled temperature difference reached zero or below; a non-converging property call was treated the same way.
A failed run stopped after three consecutive non-converging property calls and returned the penalty vector of −1,000,000 W net power and 1,000,000 kW heat input. The penalty was excluded from the reported Pareto points but counted in the failed-run total.
| Cores | Wall time | Speed-up | Failed runs |
|---|---|---|---|
| 1 | 41 min 36 s | 1.00× | 11 of 240 |
| 4 | 11 min 09 s | 3.73× | 11 of 240 |
| 8 | 6 min 01 s | 6.91× | 11 of 240 |
The final run produced 17 valid Pareto points. Three representative points were 412 kW net power with 1,245 kW heat input, 428 kW with 1,382 kW, and 447 kW with 1,612 kW. The points expose the trade-off directly: higher power required more heat input in this model.
| Pareto point | Net power | Total heat input | Decision meaning |
|---|---|---|---|
| A | 412 kW | 1,245 kW | Lower heat input for efficiency-focused designs |
| B | 428 kW | 1,382 kW | Balanced compromise for a general design study |
| C | 447 kW | 1,612 kW | Higher output when power capacity dominates |
Which EES optimization methods should you choose?
When is built-in EES optimization enough?
Use built-in EES optimization when one objective, finite bounds, and a manageable function-call budget answer the engineering question. Commercial or Professional is usually the relevant license comparison for multi-variable work.
When should you run parallel EES optimization?
Run parallel EES optimization when you need a Pareto set, repeated restarts, or a wall-time reduction across independent model evaluations. The benchmark’s 6.91× speed-up on eight cores is the useful decision signal, not the core count by itself.
Is Genetic or DIRECT the better method?
Genetic is the more robust choice when the response surface is difficult, but the EES Manual describes it as slow. DIRECT can reach global optima more efficiently when your bounds and function behavior suit it. Test both only when the added calls fit your budget.
Choose the workflow that matches your objectives
The neutral decision rule is simple: use EES’s built-in methods for bounded single-objective studies, and use parallel multi-objective execution when the design decision requires several objectives or many independent solves. For the documented `$Import`/`$Export` and `EES.exe /solve` workflow, see the ees software page.