Publications

Peer-reviewed conference and journal papers, the PhD thesis, books, and whitepapers.
Each entry links through to a page with the abstract and links to the publication, code, or related material.

For an always-current list with citation counts, see Google Scholar or the SDU Find Researcher portal.

Research keywords: Scalability · IoT · Software Engineering · Software Architecture.

The research contributions below have been carried out as part of projects funded by Innovationsfonden, EUDP, ForskEl (Energinet), the EU Regional Development Fund, and SDU.

Operationalize Self-Managing Machine Learning Systems with a Generalizable Knowledge-Driven Architecture

A generalisable Knowledge-Driven Architecture (KDA) that operationalises self-managing ML systems by embedding semantic models in a Knowledge Graph. Declarative probes drive automated integrity monitoring and MAPE-K reconfiguration, achieving 0.56 s adaptation latency and 97% consumption reliability in a synthetic manufacturing setting – a concrete fit for EU AI Act post-market oversight.

research

knowledge-driven architecture

machine learning

self-adaptive systems

software architecture

Anders Launer Bæk-Petersen, Mahyar T. Moghaddam, Jakob Hviid, Mikkel Baun Kjærgaard

AI Pipelines: A Scalable Architecture for Dynamic Data Processing

A scalable, modular AI pipeline architecture with a Heartbeat-Update-Synchronize-Ingest-and-Register (HUSIR) protocol for orchestration, fault tolerance, and dynamic job routing. Evaluated over two years in production at Esoft, the architecture replaced a manual two-week pre-processing cycle with automated, A/B-testable AI workflows and matured from TRL 3 to TRL 7.

research

AI pipelines

software architecture

distributed systems

microservices

Jakob Hviid, Anders Launer Bæk-Petersen, Emil Stubbe Kolvig-Raun, Juan Francisco Marín Vega

Balancing Cobot Productivity and Longevity Through Pre-Runtime Developer Feedback

A machine-learning approach for static code analysis of cobot programs that predicts longevity impact at compile time, before deployment. Trained on data from 1,325 Universal Robots e-Series cobots, the model classifies each program line by expected wear and reaches 90.43% worst-case accuracy. The accompanying dataset of 56,405 unique program-line executions is published for follow-up work on sustainable robotics.

research

cobot

robotics

machine learning

developer feedback

software engineering

Emil Stubbe Kolvig-Raun, Jakob Hviid, Mikkel Baun Kjærgaard, Ralph Brorsen, Peter Jacob Sørensen

On Reserve Markets in the Era of High Storage and Flexibility Penetration

Existing reserve products are mismatched to storage and flexible demand in renewables-driven systems, leaving new players exposed to systemic risk. The paper introduces Energy Reserves – energy-capacity reserves spanning timeslots – that let storage units express their full capability without the over-conservative or imprudent extremes of today’s per-slot products. Reserve costs in Denmark went from an estimated 700M DKK in 2023 to 1.6B DKK actual, making this fiscally urgent.

research

power markets

flexibility

renewables

energy reserves

Georgios Tsaousoglou, Jakob Hviid, Hanne Binder, Henrik Madsen

Driving Towards Grid Balance

A technical framework for democratising the EV charging market by separating infrastructure ownership from service operation. Charge-point owners can delegate granular control and read access to designated actors, lowering market entry barriers and enabling tighter coupling of charging behaviour to renewable production. The pattern generalises to heat pumps and the wider EU Data Space initiative – with 1.5 million EVs expected in Denmark by 2035.

whitepaper

electric vehicles

smart grid

data ecosystem

data spaces

regulation

Laura Schmidt Arenkiel, Jakob Hviid, Morten Houborg Andersen, Mikkel Siggaard Clausen

OPM: An Ontology-Based Package Manager for Building Operating Systems

An ontology-based package manager for Building Operating Systems that resolves dependencies semantically – across hardware, sensor capabilities, and the building’s physical layout. Where apt or npm can only express syntactic dependencies, OPM treats ‘needs a temperature sensor in this room’ as a first-class dependency type and verifies the building actually satisfies it before deploying a service. Result: significantly reduced deployment and maintenance time for BOS-based DR services.

research

building operating systems

ontology

package manager

smart grid

demand response

Jakob Hviid, Aslak Johansen, Fisayo Caleb Sangogboye, Mikkel Baun Kjærgaard

Service Portability and Information Discovery in Building Operating Systems using Semantic Modeling

A semantic information-discovery mechanism for Building Operating Systems that integrates with existing ontologies such as Brick. Applications query for what they need (e.g. an occupancy estimate for a specific room) rather than calling specific endpoints, so services can be merged, split, or arbitrated without code changes in dependent applications. Demonstrated by porting nine services across three real building models – a collaboration with Gabe Fierro (UC Berkeley).

research

building operating systems

ontology

service discovery

semantic modeling

Jakob Hviid, Aslak Johansen, Gabe Fierro, Mikkel Baun Kjærgaard

Experience Report: A Systematic Process For Gathering Quality Attribute Requirements for Industry 4.0 Middleware

A three-phase systematic process for gathering quality attribute requirements for I4.0 middleware from industry partners. Combines initial data collection with domain experts, a Quality Attribute Workshop (QAW), and structured Quality Attribute Scenario evaluation. Applied in a real industry-academia project group, the process yielded 6 industrial high-level requirements and 14 derived QASes mapped to 10 quality attributes – a repeatable path from business-language requirements to architecture-actionable scenarios.

research

industry 4.0

middleware

quality attributes

software architecture

requirements engineering

Sune Chung Jepsen, Torben Worm, Henrik Bærbak Christensen, Jakob Hviid, Lars Märcher Sandig

Industry 4.0 Middleware Software Architecture Interoperability Analysis

Applies the Levels of Conceptual Interoperability Model (LCIM) to analyse the implications of multiple interoperability levels in I4.0 middleware software architecture. Proposes an event-driven middleware composed of a publish-subscribe bus, asset-bridge components for REST/MQTT/OPC UA/SOAP, and a controller – with ontology-based service descriptions layered on top to handle technical, syntactic, and semantic interoperability. Companion to the SAC 2021 asset-side analysis.

research

industry 4.0

middleware

software architecture

interoperability

Sune Chung Jepsen, Torben Worm, Thomas Ingemann Mørk, Jakob Hviid

An Analysis of Asset Interoperability for I4.0 Middleware

Maps the Levels of Conceptual Interoperability Model (LCIM) onto Industry 4.0 middleware, using the MASON ontology to make each level concrete with a real I4.0 lab case. Argues that most existing I4.0 middleware focuses on the technical and syntactic levels, while the pragmatic and dynamic levels – where exchange meaning depends on context and systems negotiate behaviour at runtime – are where flexible production actually lives. Foundation paper for several follow-up middleware analyses.

research

industry 4.0

middleware

interoperability

ontology

Sune Chung Jepsen, Torben Worm, Thomas Ingemann Mørk, Jakob Hviid

A Pilot Study of Industry 4.0 Asset Interoperability Challenges in an Industry 4.0 Laboratory

Pilot study at SDU’s 18.5M EUR Industry 4.0 laboratory, conducting asset integration as part of building an Information Backbone – a software infrastructure that integrates warehouse, transport, and robotic systems. The work reveals that asset interoperability readiness varies wildly across vendors: missing external interfaces, poor documentation, and inconsistent technologies. Becomes the running case study for several subsequent LCIM-based interoperability papers.

research

industry 4.0

interoperability

pilot study

middleware

Sune Chung Jepsen, Thomas Ingemann Mørk, Jakob Hviid, Torben Worm

Modeling and Performance Simulation of a Retail Store as a Smart Grid Ready Building

A detailed EnergyPlus simulation of a Danish supermarket, evaluating flexibility scenarios with the building’s actual work processes folded in via an ethnographic study. Most building-flexibility studies model loads as if buildings ran themselves – here the freezer cycle, oven cycle, and ventilation cycle are tied to what staff actually do (restocking, bake-fresh-bread schedules, cleaning routines). Result: more realistic flexibility numbers and a method other retail studies can borrow.

research

smart grid

retail

building simulation

flexibility

Muhyiddine Jradi, Henrik Engelbrecht Foldager, Rasmus Camillus Jeppesen, Jakob Hviid, Mikkel Ask Rasmussen, Mikkel Baun Kjærgaard

A Software Approach to Mitigating Barriers for Smart Grid Integration in the Retail Sector

PhD thesis on the software-architecture barriers to integrating retail buildings into smart grids. Proposes a layered approach: an empirical survey of retail loads and flexibility, an Activity-Tracking Service for context-aware Demand Response, a Service Abstraction Layer for portability and resilience, auto-configuring services to cut deployment cost, and an ontology-based approach to information discovery. Each contribution is grounded in a peer-reviewed paper from real retail and lab settings.

thesis

smart grid

retail

building operating systems

software architecture

Enabling Auto-Configuring Building Services: The Road to Affordable Portable Applications for Smart Grid Integration

Tooling that brings auto-configuration to Building Operating Systems, supporting service discovery and publication at runtime with only minor changes to existing services. The aim is to make the marginal cost of adding a Demand Response service close to zero, shifting integration complexity from per-site entrepreneurs to application developers who can amortise it across many buildings. Critical for hitting the ~3-year ROI horizon Danish retail expects on building investments.

research

building operating systems

smart grid

service discovery

demand response

Jakob Hviid, Aslak Johansen, Fisayo Caleb Sangogboye, Mikkel Baun Kjærgaard

Service Abstraction Layer for Building Operating Systems: Enabling portable applications and improving system resilience

A Service Abstraction Layer (SAL) for Building Operating Systems that decouples applications from specific service implementations. Applications no longer call ‘/api/v1/occupancy on host x’ – they request a capability and the SAL routes them to a service that provides it, enabling provider swaps without code changes and parallel providers for resilience. The conceptual foundation that the 2022 ontology-driven Service Portability paper builds on.

research

building operating systems

service abstraction

smart grid

resilience

demand response

Jakob Hviid, Mikkel Baun Kjærgaard

The Retail Store as a Smart Grid Ready Building: Current Practice and Future Potentials

A bottom-up survey of four Danish retail store types – hypermarket, supermarket, garden centre, hard goods – documenting actual loads, flexibility potential, and current control granularity. Ovens take up 60% of supermarket flexibility potential and 19% of hypermarket potential, but the existing controllable devices are too coarse to be used in modern Demand Response. The empirical foundation for the rest of Jakob’s PhD work, which then attacks the abstraction and deployment-cost barriers.

research

retail

smart grid

demand response

flexibility

Jakob Hviid, Mikkel Baun Kjærgaard

Activity-Tracking Service for Building Operating Systems

An Activity-Tracking Service (ATS) for Building Operating Systems, adding an explicit activity layer on top of existing BOS proposals like Brick, BOSS, XBOS, and Bosswave. Lets Demand Response applications reason about why a load is on, not just that it is on – so a refrigeration cycle triggered by just-restocked warm products is treated differently from a scheduled cycle that could safely shift. Designed around BOS security, privacy, and scalability constraints.

research

building operating systems

IoT

activity tracking

retail

demand response

Jakob Hviid, Mikkel Baun Kjærgaard

Clustering and Visualisation of Electricity Data to Identify Demand Response Opportunities

A poster-abstract data-driven method that clusters and visualises raw building electricity data to produce day-type profiles – winter weekdays, weekends/holidays, and seasonal variants. Pre-processing handles summer/winter timestamp artefacts, multivariate outlier detection cleans the data, and k-medoids on 24-feature daily vectors with silhouette-width selection produces the clusters. Used as a screening step to rank a portfolio of retail sites by ‘how worth-auditing they look from meter data alone’ in the FlexReStore project.

research

clustering

demand response

electricity data

retail

Almir Mehanovic, Emil Sebastian Rømer, Jakob Hviid, Mikkel Baun Kjærgaard