Comparison of Energy System Optimization Software and Evaluation of Selected Frameworks
Source: arXiv:2607.16121 · Published 2026-07-17 · By Pedro Caixeta, David Gawron, Hüseyin K. Çakmak, Haozhen Cheng
TL;DR
This paper addresses the challenge of selecting suitable energy system optimization software among multiple available tools, each targeting different optimization goals and user needs. The authors perform a detailed comparative study of five frameworks—REMix, MTRESS, COMANDO, OEMOF, and HOMER PRO—by conducting an extensive literature review, examining tutorial models, and developing criteria for systematic evaluation. Due to insufficient documentation and runnable examples, MTRESS was excluded from the full comparison.
The analysis reveals distinct strengths for each tool: REMix is specialized for investment path optimization over long-term horizons in large energy systems; COMANDO excels in modeling highly detailed systems with unconventional components; OEMOF stands out as the most user-friendly open-source option suitable for standard optimization problems; and HOMER PRO, with its graphical interface, enables rapid synthesis and design optimization well suited for users with limited programming skills. The authors highlight trade-offs in financial cost, training effort, modeling flexibility, model complexity, and supported optimization goals. Overall, this work provides practical guidance for choosing the most appropriate software based on project requirements and user expertise.
Key findings
- REMix requires a GAMS license costing approximately $3500 for full functionality; academic discounts up to 80% apply.
- OEMOF and COMANDO are free and open-source but depend on external solvers, some of which may require licensing.
- HOMER PRO licenses range monthly from $9 (student) to $380 (standard), with free trials available.
- Training effort varies: 1 day for REMix and OEMOF for experienced Python users; 2 days estimated for COMANDO due to higher complexity; HOMER PRO requires minimal programming knowledge and can be learned in ~1 day.
- Modeling effort is lowest in HOMER PRO due to extensive preset components and built-in access to location-specific data; COMANDO has highest modeling effort since users build components from scratch in Python.
- REMix supports investment path optimization (optimal timing of investments), which was not found in other tools.
- COMANDO supports the widest range of customizable components, including thermal masses and complex industrial elements, allowing modeling of atypical components.
- HOMER PRO supports up to 56 components per model with limits on repetition; REMix, COMANDO, and OEMOF have no known limits and can handle models with over 100 components.
- Optimization goals differ: REMix and HOMER PRO focus on economic cost minimization; OEMOF can handle economic, environmental, dispatch, and net power maximization; COMANDO examples cover economic and power maximization.
- HOMER PRO automates evaluation of multiple design variants and sensitivity analyses, facilitating design and synthesis optimization.
- Difficulties running tutorials due to licensing and solver issues limited hands-on evaluation, especially for REMix and COMANDO.
Methodology — deep read
The authors structured their methodology in two phases. Initially, a comprehensive literature review was conducted, gathering information on five energy system optimization tools (REMix, MTRESS, COMANDO, OEMOF, HOMER PRO) including their goals, tutorials, license costs, and interfaces. This review highlighted a lack of sufficient usable examples for MTRESS, leading to its exclusion from in-depth comparison.
In the second phase, the authors analyzed a total of 62 tutorial models from the remaining four tools. They examined model components, interconnections, and stated optimization objectives. Based on patterns and differences observed, they iteratively developed a set of six comparison criteria divided into Effort (financial cost, training effort, modeling effort) and Usage Scope (optimization goal, optimization object, modeling scope).
Each software was evaluated against these criteria by detailed examination of official tutorials, documentation, and literature examples; for instance, they noted REMix’s unique investment path optimization capability, HOMER PRO's GUI and variant analysis, and COMANDO’s component flexibility. The comparison includes qualitative assessments supported by counts (e.g., number of tutorials, component limits) and cost figures (e.g., GAMS license prices).
Training effort was gauged from tutorial availability, example complexity, and programming requirements, especially Python skill demands. Modeling effort considered user interface, code volume, and embedded data availability for components.
Due to project time constraints and dependencies on proprietary software (GAMS for REMix), direct execution and benchmarking of models were limited or not possible, affecting empirical validation. No statistical tests or cross-validation were performed; evaluation relied on qualitative and literature-sourced insights.
Reproducibility is constrained: no code implementations were created by authors; documentation and materials used are publicly available except MTRESS examples, which were incomplete. Weighted scoring or numeric ranking across criteria was not reported.
Overall, the methodology is a structured, literature and tutorial-driven qualitative comparison emphasizing practical selection guidance rather than empirical benchmarking or performance metrics.
Technical innovations
- Introduction of six structured comparison criteria combining effort and usage scope factors for energy system optimization software evaluation.
- Highlighting and clarifying the unique investment path optimization feature of REMix among benchmarked tools.
- Demonstrating the usability spectrum from code-intensive, component-flexible software (COMANDO) to GUI-driven rapid synthesis tools (HOMER PRO).
- Emphasizing modeling scope differences in energy transmission types, component variety, and model size limits across tools.
Baselines vs proposed
- REMix: requires licensed GAMS solver (~$3500) vs OEMOF and COMANDO: free open-source with potential free solvers.
- Training effort (experienced user): REMix and OEMOF ~1 day vs COMANDO ~2 days vs HOMER PRO <1 day.
- Model size limit: HOMER PRO ≤ 56 components vs REMix >100 components (no known limit).
- Optimization goals frequency: HOMER PRO economic optimization in 20 of 21 tutorials vs OEMOF dispatch optimization in 13 of 20 tutorials.
Figures from the paper
Figures are reproduced from the source paper for academic discussion. Original copyright: the paper authors. See arXiv:2607.16121.
Fig 1: Schematic representation of our work process to develop comparison criteria.
Fig 2: Cathegorization of the comparison criteria. The blue boxes mark the 6 final criteria.
Limitations
- Unable to obtain adequate usable documentation and working examples for MTRESS, excluding it from detailed comparison.
- Could not execute REMix tutorials due to GAMS licensing restrictions, limiting practical evaluation.
- Installation and solver issues prevented running four COMANDO tutorials, hindering empirical validation.
- Comparisons largely qualitative and based on documentation and tutorials, lacking empirical model performance benchmarks.
- Modeling effort assessment is subjective without implementation or timed user studies.
- Limited project time (~180 hours) constrained depth of analysis and coverage.
Open questions / follow-ons
- How do these software tools compare under real-world large-scale case studies with measurable performance metrics such as runtime and solution quality?
- Can MTRESS be more fully evaluated once documentation and example issues are resolved, and how does it compare to other tools?
- What improvements in solver integration and licensing models could enhance usability and accessibility of proprietary-dependent tools like REMix?
- How can modeling effort be objectively quantified across tools through user studies or controlled experiments?
Why it matters for bot defense
Though this paper focuses on energy system optimization rather than bot defense or CAPTCHA technologies, there are indirect lessons for bot-defense engineers in software evaluation methodology. The authors demonstrate a systematic approach to software tool comparison incorporating licensing costs, usability, training, modeling scope, and optimization goals—criteria analogous to evaluating tools for CAPTCHA or bot mitigation tasks.
For practitioners in bot-defense or CAPTCHA, this study highlights the importance of aligning tool selection with the specific problem context and user expertise. For instance, just as HOMER PRO suits quick prototyping with a GUI for non-expert users, bot-defense engineers might prefer tools that balance ease-of-use and configuration over highly customizable but complex frameworks depending on project constraints. The observed tradeoffs between modeling effort and flexibility resonate across defensive technology choices. However, since no adversarial or security evaluation was conducted, direct insights into robustness assessment or attack simulation valuation remain outside this paper's scope.
Cite
@article{arxiv2607_16121,
title={ Comparison of Energy System Optimization Software and Evaluation of Selected Frameworks },
author={ Pedro Caixeta and David Gawron and Hüseyin K. Çakmak and Haozhen Cheng },
journal={arXiv preprint arXiv:2607.16121},
year={ 2026 },
url={https://arxiv.org/abs/2607.16121}
}