# 13. Conclusions¶

## 13.1. Technology Development¶

The technology development in Annex 60 was organized around three standards:

1. IFC for data modeling,
2. Modelica for multi-domain, multi-physics modeling, and
3. FMI for run-time interoperability of simulators.

Basing the work on these standards was critically important to enable joint technology development among multiple participants, many of whom brought into Annex 60 their existing code, further developed it within the project, and then integrated it back into their tools. A prime example of technology developed around these standards is the Modelica Annex60 library developed in Activity 1.1. In this work, multiple institutes started a collaborative development of a free, open-source Modelica library for building and district energy systems. This work harmonized the previously fragmented and duplicative development of libraries, and resulted in a jointly developed library that is now used by four major Modelica libraries for building systems. As part of this process, the four libraries not only grew in their functionality, but also improved their robustness, validation and documentation. Getting to this point required the developers of previous libraries, each having a considerable code base, to mutually agree upon common processes for development and quality control. These processes needed to allow for rapid experimentation, as is often done in University settings, as well as ensure robust and stable development, which is more important for commercial software companies and government laboratories. The joint development also required the developers to agree upon various design decisions and conventions for coding, documentation and validation, to jointly work on implementation and vetting of a core of a library, to refactor their existing libraries, and to open-source previously proprietary code. This is the first international collaboration for a library with free, open-source models for buildings and district energy systems that are built using an open-standard modeling language. It initiated a larger open-source development that will be further supported though IBPSA, whose vision includes providing a standard library with fundamental model descriptions that will be supported by manufacturers and integrated in various building performance simulators.

A second example of technology developed using the previously mentioned standards is the development of building and district energy simulation tools based on the FMI standard. A key advantage of using the FMI interface is that it decouples the model authoring from the simulation run-time environment, allowing quite different approaches to be used to integrate the models in time. Over the course of the project, different FMI simulators not only emerged from within Annex 60, but also from within the significantly larger FMI community. Within Annex 60, using the FMI standard as an Application Programming Interface (API) to different simulators allowed run-time coupling of various simulators, prototyping of an interface that allows energy and control models authored in FMI for co-simulation to be integrated with a commercial building automation system for real-time control or real-time energy monitoring, and hardware-in-the-loop testing of building control sequences. From the technology development point of view, such a decoupling between model authoring and simulation, and work with an open standard, has the key advantage of being able to develop, prototype, and test different simulators on common models, while keeping the time and cost to adapt a new simulator low. This is compared to what is done in most modeling and simulation tools, which have a tight integration of models and solvers. While it has been shown that the FMI standard is well-suited for the classes of problems encountered in building and district energy simulation, certain limitations were encountered, as is common with any standard that covers such a complex use case, and improvements have been recommended. Some of these recommendations have already been addressed by the FMI standards committee. Overall, using the well-developed FMI standard, together with its development process that ensures stability, gave rise to the ecosystem of building and district energy simulation tools developed as part of this project.

A third example of technology development using these open standards is that which was done to support processes that transform digital planning and design to simulation. In Annex 60, a mechanism was developed to transform a digital model of a building and its energy systems to Modelica code, which can then be readily used for advanced building performance simulation. This was accomplished through the use and extension of the Open BIM data formats defined by the Industry Foundation Classes (IFC) as well as through the use of other BIM-standards, such as the Information Delivery Manual (IDM) and Model View Definitions (MVD). Annex 60 thoroughly addressed the prevailing tedious, cumbersome and error-prone process of manual data conversion and model generation by providing a methodology and software framework for automatically, or at least semi-automatically, transforming a digital model into an object-oriented acausal model. The software framework developed in Annex 60 supports the Open BIM format IFC. Models are checked for integrity in terms of geometric consistency and HVAC definition. Using a flexible module for schema parsing, models are transformed into the intermediate data format SimXML. To manage these SimXML data, a dynamic schema parser in C++ with an API between C++ and Python was developed to interact with the data model. An object and parameter mapping mechanism as well as respective mapping rules were defined to formulate engineering knowledge in a rule-based methodology. These rules are processed by the framework. Furthermore, an IFC MVD was published in order to specify the subset of IFC data relevant to building performance simulation. To support multiple Modelica libraries, we selected a template-based approach and implemented it in Python. As a result, the software framework of the translator from BIM to Modelica can generate models using the four different Modelica libraries, using the same software stack with small adaptations. The developed methodology was tested for a set of use cases. Annex 60 thereby followed a bottom-up approach which was based on these use cases. The framework supports the whole model checking and model transformation process but is currently limited to these use cases. The result is a modular framework which is open for further development and dissemination.

In addition to the use of open standards, workflow automation also played an important role in technology development, both for model development and also for case studies. Due to the large ecosystem of free, open-source Python packages, Python-based tools were further developed and used in the case studies, and documentation and workshops for novice users were developed. An example for workflow automation during model development is the regression testing that we setup for the Annex60 Modelica library. This library consists of $$2,400$$ files. During any change to a model, more than $$120,000$$ result points are compared to verify that they are either identical to or expectedly different from results computed with an earlier version, within a tolerance of $$10^{-3}$$. Furthermore, daily tests are run to compare all of these points among the two simulators Dymola and JModelica to make sure that the library produces the same result with different simulators. Regarding workflow automation for case studies, many case studies used optimization, stochastic simulations, or coupled multiple simulation tools. For such applications, automating the workflow was critical to produce reliable and repeatable results.

Subtask 2 demonstrated how these technologies can be used for the design and operation of building and district energy systems. In many of these applications, models from Activity 1.1 were combined with FMI-tools from Activity 1.2 and workflow automation scripts from Activity 1.4, and with open-source and commercial software that has been developed outside of Annex 60, to solve problems related to the design and operation of building and district energy systems. The case studies showed that very low energy systems and increased grid integration imposes structural changes to building and community energy modeling and simulation tools and processes. For example,

• Models of different physical domains, control systems, occupant behavior, and occupancy need to be combined for dynamic, multi-physics simulations that involve electrical systems, thermal systems, flow distribution, controls and possibly communication systems. Such models typically evolve at vastly different time scales.
• Tools need to properly handle control sequences and hybrid systems in which the states evolve in time based on both continuous and discrete time semantics that arise from physics and digital control, respectively.
• Subsystem models need to be extractable in order to export and execute models in a self-contained form in a building automation system, or a hardware-in-the-loop setup.
• Model equations need to be accessible in order to perform model order reduction, model linearization, extraction of a linear model in state-space representation, and to solve optimal control problems.

The case studies showed Modelica is well-suited to address the above problems. Various case studies combined thermal, electrical and control models for projects at the building and district scales. While Modelica tools generate very efficient C-code for simulation, simulating large systems can still be a challenge. Research in translation of large models and in multi-rate solvers with adaptive time step control is ongoing and needed to apply this technology to large systems.

As part of analyzing thermal district energy systems, it becomes evident that a standardized validation test, similar to ANSI/ASHRAE BESTEST 140, but for district systems, would be immensely valuable. Currently, there is neither a test, nor a common understanding of what is meant by a district energy simulation. Approaches in industry range from simple spread-sheet based analysis that disregard temperature levels, energy storage and pressure drops, to fully coupled simulation of buildings, thermal distribution networks, flow friction and feedback control. In absence of a test procedure for district models, it is hard to judge how credible the tool and approach that was selected by the modeler is. A first start for such a test was done in Annex 60, and it is planned to continue this effort within the IBPSA Project 1.

Annex 60 participants who worked on model use during operation reported that the object-orientation of Modelica not only enhances reusability of the models, but also allowed models to be developed following the physical system structure, which makes them easier to understand. Moreover, the possibility to import and export Modelica models as FMUs enabled the integration of models using a standardized, tool-independent API into independent or combined solutions for data analysis, simulation, fault detection and diagnosis, and optimization now and in the future.

In summary, had our work not been based on such open standards, such a tight collaboration and integration of software among widely ranging use cases would not have been possible. Moreover, analyzing multi-physics problems such as the integration of thermal and electrical systems, overlaid with controls that coordinates the two, would have been very difficult without Modelica’s support for multi-physics modeling, and FMI’s support for co-simulation using domain-specific tools.

## 13.2. Governance and Dissemination¶

The Annex 60 benefited from a strong organizational framework to coordinate work between the various researchers world-wide. Eight semi-annual international expert meetings were conducted, most of them followed by technical workshops over multiple days. To synchronize work packages and activities, more than 120 online coordination meetings and web-conferences were held within the activities and among activity leaders. A Bitbucket repository system was used, including a wiki, as a key resource to organize collaboration and manage shared documents. Several special scientific tracks at national and international conferences were organized to disseminate and promote results of the entire project, which helped to gain a high international visibility. Annex 60 published 11 journal articles, 38 conference papers, a Modelica library and diverse source code packages, all managed through a Git code repository.

## 13.3. Continuation Framework¶

Although the Annex has delivered a solid basis of tools and demonstrated their application, tool development is a continuous effort. It continuously needs to provide support, respond to new system technologies and apply advances in computer science and applied mathematics. Upcoming research and development issues that remain to be solved after completion of the Annex will therefore be coordinated under the umbrella of the network of the International Building Performance Simulation Association (IBPSA) as a living and supporting dissemination framework.