Image: GPLv2+ license notice for the now‑defunct deeco energy system model added on 24 February 2003. Courtesy Robbie Morrison
The Open Energy Modelling Initiative (shortened to openmod) is an online and offline umbrella community devoted to promoting open energy system modeling and analysis. While there are no restrictions on application area, the bulk of funded research is directed toward questions involving public policy. As of late‑2019, the openmod has about 600 participants on its mailing list, with most of them being full‑time researchers or analysts. More information on Wikipedia. Key URLs are listed in the infobox below.
Open Energy Modelling Initiative
Services are publicly readable by anyone but registration is required to post.
Mailing list – https://groups.google.com/forum/#!forum/openmod-initiative
Forum – https://forum.openmod-initiative.org
Wiki – http://wiki.openmod-initiative.org
Overview – https://forum.openmod-initiative.org/t/1032
Robbie Morrison began building energy system models in 1995. He released the first open source energy model in 2003 under a copyleft GPLv2+ license (see screenshot). Since that point, he has advocated that all models used to inform public policy be open in the interests of democracy. More recently, he has been promoting the idea that the informed public must be involved in designing future energy system scenarios. His current project is to initiate an online community, drawn from civil society, to undertake the open analysis of energy and climate policy to compliment and counter official findings.
Robbie is not present on social media so if you wish to contact him please email robbie.morrison at his provider posteo.de.
This posting describes the openmod community — its history, structure, governance, ethos, activities, challenges, and outlook. But first some key thoughts:
- numerical modeling is a necessary part of understanding energy systems
- genuine open analysis can deliver much needed public policy transparency
- online communities enable diverse participation and retain soft knowledge well
- periodic workshops — in person or via video — can compliment online activities
- community‑curated numerical data is a core resource for all
- open analysis has the potential to recast public policy making in the Global South
- governance and infrastructure refined by three decades of online open source software development can and should be leveraged in the overlapping interests of reproducible science and open policy analysis
Energy system modeling
Let’s look briefly at the domain itself, starting with a topical example. Adding a hydrogen sector to the energy mix offers many potential advantages and is an idea moving rapidly up the policy agenda. However the benefits and drawbacks are not easily quantified and, as argued here, are best revealed using computer-based experiments.
Energy system modeling involves coding suitable tools, collecting numerical data, and running simulations of future energy systems to assess their feasibility and then explore their other attributes. The “open” qualifier means that all aspects of the modeling pipeline are legally and technically open, including the source code for the tools, the data itself, and the documentation. The workflow should ideally be scripted and easily reproducible as well.
The reason for this often painstaking numerical approach is that energy systems are simultaneously complicated (necessitating appropriate detail) and complex (exhibiting emergent behavior). There are few if any self‑evident or intuitive answers to most public policy questions, particularly those concerning rapid and complete decarbonization.
The most common (but not the only) modeling paradigm is to represent the underlying structure and processes as (excuse the jargon) capacitated graph‑dynamical systems — such that the resulting models are quite literal in form and high‑resolution in time (hourly or sub‑hourly), space, and network detail (topology). Given a scenario is feasible, the resulting solution space needs to be somehow navigated to yield a single trajectory through time to some future carbon neutral year (typically 2050 at the time of writing). One widely used strategy is to minimize the aggregate financial cost under perfect foresight.
Because future contexts (such as commodity prices, technology costs, engineering developments, and social change) cannot be predicted, energy system modelers necessarily fall back to scenario analysis. Under scenario analysis, different narratives about the future (known as storylines) are first developed, then quantified, and then converted to model inputs for simulation and analysis. One scenario (typically some estimate of business‑as‑usual) is selected as a reference against which other scenarios are compared by difference.
One reason for modeling the entire decarbonization horizon (out to 2050) and not just the next few steps is that potentially undesirable lock‑in/out effects, and other counteractive responses, as well as beneficial path dependencies, and other synergies, can more readily be represented and investigated. The repercussions of both poor and astute system architectural, structural, and even operational choices made today can potentially persist for decades.
In September 2014, around 30 like‑minded researchers met in Berlin and formed the openmod (I wasn’t present but joined the mailing list 18 months later). As noted, the openmod is an umbrella community spanning a number of projects, disciplines, and institutes, as well as the interested public — indeed anyone can join and contribute. I now want to talk about how that community works — the social dimensions of collaboration.
The first thing to note is the lack of structure. This community is no more than a collection of online services, primarily an email list, a discussion platform, and a wiki. There is no membership, no incorporation under law, no fees, no bank account, and no charter. But there are nonetheless unwritten rules which have arisen naturally through discussion. The first tenet is that the openmod collectively does not hold positions, nor does it endorse individual projects or initiatives. That said, members are welcome to use community infrastructure to canvass support for new ideas and nascent projects. The second tenet is that whoever organizes one of our regular workshops (described shortly) have complete dominion — benign albeit temporary dictatorship so to speak. What little money is required (for example, annual domain name registration fees or server space) is kindly donated by individuals.
The community alternates between these continuous online activities and occasional workshops. The workshops are not dissimilar to the first and ten have been held at the time of writing. These workshops, normally limited to 60 modelers, compliment the online activity and allow community members to meet in person. They are explicitly hands‑on to tackle and advance issues of common cause across the community. Planning and canvassing for these so‑called “do‑a‑thons” normally takes place on the openmod discussion forum.
Central European countries like Germany and Denmark are over‑represented within the openmod, mostly through context (including funder attitude) and critical mass effects. But in September 2019, the first North American workshop was held at NREL, Colorado, USA. There are comparatively few researchers from former eastern Europe, India, and the Global South generally. Geographical diversity is an issue that needs addressing as the community matures. To assist, several contributors from the community recently sought European funding to aid researchers from target‑inclusive countries.
Not surprisingly perhaps, much of the ethic underlying the openmod derives from the open source software world with its three decades of experience in managing similar communities in related spaces.
Displaced‑in‑time (asynchronous) communication is the mainstay of online communities. It allows for participation across different time‑zones and diverse daily schedules. It is also largely self‑journaling and later searchable — to become in effect the collective memory of that community. Selecting the appropriate channels and determining their public visibility can therefore be crucial to success. Choosing between self‑hosted software or free‑of‑charge cloud services may also be significant, not least in terms of managing personal data privacy.
The openmod currently offers a public mailing list (Google Groups), a forum (self‑hosted running Discourse), and a wiki (running MediaWiki), but has yet to add a chat platform or file server. While most of the traffic passes through the mailing list, the forum (which I help administer) provides a more enduring record. Figure 1 indicates potential channels and their characteristics. The openmod has elected not to operate registration-for‑read services.
Figure 1 : Asynchronous communication channels spanning the more transient to the more lasting. The channels used by the openmod community are indicated by the gray block. The selected constellation of services would normally evolve with both community needs and developments in communications generally.
Online communities tend to retain knowledge well, particularly the soft knowledge held by individuals. Online communities thus allow researchers to remain in contact with their projects and research fields long after their formal involvements have ended — often an extremely valuable resource that would otherwise have evaporated.
Governance and conduct are always issues for communities, be they online or offline. The openmod has neither a code of conduct nor a formalized disputes process. To my knowledge, the openmod has never faced a problem that could not be resolved with a single email or a chat in a corridor. It is my firm belief that any code of conduct, whether aspirational or indicating minimum standards of behavior, must be accompanied by a clearly defined and articulated disputes process. In the absence of a code of conduct, the openmod takes the view that the moderators on each of the communication platforms are solely responsible for conduct on their particular platform. On balance, that strategy has worked just fine so far, although the past does not necessarily guarantee the future.
Open data as social glue
I often think of data as the glue that binds numerically‑intensive research communities — the single area where agreement and collaboration pays off the most. Unlike code, data provenance and interoperability are paramount. Meeting these two criteria requires accurate metadata and a broad agreement on technical, legal, and semantic standards. The term “standard” is used here loosely in the sense of common practice within a community, whether explicitly caucused and consented to or not.
Technical interoperability is often achieved by settling on some suitable common denominator. In this context, a number of energy system research groups have opted for the Open Knowledge Foundation (OKFN) tabular data package based on character encodings (as opposed to more space efficient binary encodings). Figure 2 shows how such data packages, with containing metadata, can be structured. Work continues within the openmod community on developing common technical standards for datasets, data packages, and metadata.
Figure 2 : Structure of a generic OKFN data package, depicted as a UML class diagram. The metadata is bundled with the package and would include an SPDX identifier for the particular open license that applies.
Legal interoperability requires the uses of data‑capable open licenses (Ball 2014). But in order to avoid strictly open but nonetheless legally‑incompatible data silos, it is now recommended that data providers release only under a Creative Commons CC‑BY‑4.0 license or a CC0‑1.0 public domain grant at their discretion (Lämmerhirt 2017).
Semantic interoperability also necessitates common practice. The Open Energy Ontology (OEO) project, which commenced in September 2019, will develop and hopefully agree a set of concepts and terminologies in tandem with this community and other stakeholders.
Unlike other areas of science, energy modelers are heavily dependent on what the European Commission terms “privately held information [of] public interest”. In an attempt to counter market failure and aid system security, much of this data is now subject to statutory publication but not mandatory licensing. So although this information is available on the internet, it cannot be modified (reformatted, corrected, extended, and/or combined) and re‑published with legal certainty — a state of affairs that impedes community curation and more generally prevents improvements to data quality and utility.
As a consequence, members of the community have been active in promoting open data licensing. Some 39 individuals collectively submitted their views to the recent reform of public sector information (Morrison et al 2017). Other participants have pushed for energy sector information providers to add open licenses, including transmission system operators via ENTSO‑E, government regulators such as BNetzA, and the European Commission. Steady progress is being made.
There are notable community data‑related initiatives. The OpenEnergy Platform (OEP) project provides backend support for energy modeling, including a versioned data repository and features for managing scenarios. The Open Power System Data (OPSD) project collects, collates, repairs, and documents European electricity sector datasets in a consistent manner and serves these datasets back to both the community and anyone else who wishes to make use of electricity system information for whatever reason. The OPSD offers several download formats, but the tabular data package (indicated earlier) remains the most popular.
More recently, several scientific projects have been funded to provide sophisticated collaborative infrastructure. The Spine project is developing high‑level data storage with translators to interface with diverse data sources and different models. The SENTINEL project aims to articulate a portable standard data format to improve interoperability between open models and, in particular, to facilitate soft‑linking. The openENTRANCE project will develop a tailored collaboration platform, which doubtless means addressing a range of social and technical challenges.
Open energy system modeling has arguably arrived at the point where any numerically fluent person can download the necessary tools and datasets, design scenarios, and undertake analysis. Since November 2018 therefore, some in the openmod (myself included) have sought to partner with various stakeholders as part of the bridge project. Established NGOs in Europe were approached first but were not especially interested. Discussions are continuing with citizens’ assembly advocates. And there is now an idea on the table to form an open analysis community, distinct from the openmod itself, to further such analysis (Morrison 2019).
Global South and open analysis
Open analysis offers clear benefits to public institutions and civil society organizations that face financial constraints. Indeed, many key public policy analysis communities are constituted as excludable economic clubs with five figure annual membership fees.1 So teaming up with nascent open analysis project communities can now present a viable alternative.
Open energy analysis is now being applied to countries and regions within the Global South. In terms of energy system models, OSeMOSYS has significant uptake in Africa, South America, and other regions. OSeMOSYS underpins the pan‑African TEMBA and pan‑South American SAMBA model bases and has officially been adopted by Bolivia for its NDC calculations. Calliope has been applied in India and Brazil, Dispa‑SET in Africa and Bolivia, energyRt in India, and Balmorel in South Africa.
The energy system paradigm described at the outset can naturally be extended to cover land and water usage — an approach known variously as nexus (the water‑energy-food‑climate nexus) and CLEWS (climate, land including food, energy, and water systems) integrated assessment. Open models in this realm include GENeSYS‑MOD, MESSAGEix, NEST, and WHAT‑IF.
The electricity services planning toolkit OnSSET is used widely in Africa for assessing electrification potentials.
Established projects not only offer freely available modeling software, which anyone can download and run (with limitations described shortly), they invariably provide low‑barrier developer and user communities (much like the openmod but project‑based) and normally also ready access to the core team.
Some open models use proprietary languages (such as GAMS) and/or require or perform better with commercial optimization solvers, both of which may involve purchasing end‑user licenses for four‑figure sums. There is no real work‑around to the solver question, open source solvers are less stable, at least an order of magnitude slower, and require more memory.
Data coverage, quality, and licensing remain significant issues for the Global South. Two data portals designed to address these shortcomings specifically are the World Bank‑sponsored energydata.info and the Energy Research Data Portal for South Africa platforms.
The openmod is not the only network in this space but it is the only one specifically dedicated to “open”. Neither was it the first network for energy modeling — the Stanford‑based Energy Modeling Forum (EMF) was established soon after the 1973 oil crisis and continues to this day.
The openmod community appear to be an early adopter on several fronts — although driven as much by necessity as by opportunity, I would suggest. Milestones include the formation of the community in 2014, a collective focus on open and interoperable data, concerted efforts to persuade key data providers to apply suitable licensing, and the nascent outreach activities mentioned previously.
The outlook for the openmod appears good, particularly when set against the backdrop of steady progress on open science in Europe and elsewhere. Ongoing challenges include improving diversity and reach, experimenting with nearly carbon‑neutral conferencing (NCNC) (Hiltner 2018) and similar, reaching out to like‑minded communities in other domains, and making progress on curated data and collaboration more generally.
For specific background, search the openmod forum by category, tag, or text string.
This posting was discussed on the openmod mailing list prior to publication here. It nonetheless remains my personal view. I’d like to record my appreciation to those who kindly provided feedback (BE, DO, GL, MH, SP).
Ball, Alex (17 July 2014). How to license research data. Edinburgh, United Kingdom: Digital Curation Centre (DCC).
Hiltner, Ken (2018). A nearly carbon-neutral conference model — White paper/practical guide. Ken Hiltner website. Santa Barbara, California, USA.
Lämmerhirt, Danny (December 2017). Avoiding data use silos: how governments can simplify the open licensing landscape. Open Knowledge International. Cambridge, United Kingdom.
Morrison, Robbie, Tom Brown, and Matteo De Felice (10 December 2017). Submission on the re-use of public sector information: with an emphasis on energy system datasets — Release 09. Berlin, Germany. Creative Commons CC‑BY‑4.0 license.
Morrison, Robbie (30 September 2019). An online community for open energy analysis: improving trust, legitimacy, and participation — Release 02. Poster for EMP‑E 2019 meeting, Brussels, Belgium. Creative Commons CC‑BY‑4.0 license.
- Policy analysis clubs which overlap with energy systems modeling include GTAP, IEA, MARKAL, and WEC.
Copyright © 2019 Robbie Morrison. Text and diagrams licensed under Creative Commons CC BY 4.0 International Licenses.