The Device Chronicle

How 1KOMMA5° Orchestrates IoT Gateways for a Scalable Virtual Power Plant (VPP)

Written by Simon Buchholz | Jun 26, 2026 4:10:03 PM

The Device Chronicle interviews senior leadership at 1KOMMA5°, an Energy Management leader based in Germany. The company is rapidly growing with cleantech advancements in Europe and Australia.  



Simon Buchholz,                       Arthur Annebicque,
 Senior Product Manager      
Senior IoT Platform Engineer

With their Heartbeat AI platform, the German cleantech company 1KOMMA5° supports the renewable energy transition by connecting residential energy devices, such as solar panels, batteries, EV chargers, and heat pumps, into a virtual power plant (VPP). Heartbeat AI is an energy optimization platform linking thousands of residential edge gateways across Europe into a unified, clean energy ecosystem.

To see how this infrastructure operates reliably, Device Chronicle interviewed Simon Buchholz, Senior IoT Platform Engineer and Arthur Annebicque, Senior Product Manager of the IoT Platform team at 1KOMMA5°. Their engineering philosophy treats reliability as a visible product feature so that high cloud availability and safe edge behaviour reinforce each other.

Duck Curve – the grid problem behind a VPP

The primary purpose of the IoT Gateway within Heartbeat AI is to solve the Duck Curve – a big supply-and-demand grid imbalance common in energy markets with a high share of solar.

During peak-sun hours, photovoltaic arrays flood regional grids with energy, driving power prices down. As the sun sets, solar generation drops off rapidly just as residential consumption spikes from people returning home to charge vehicles and switch on home appliances. This creates a steep evening ramp that slow-moving traditional power plants struggle to accommodate. It is this inherent lack of elasticity in traditional power generation that slows the further integration of renewable energy.

"Heartbeat AI turns isolated homes into a unified virtual power plant," Arthur explains. "When there's a midday energy surplus, for example, our gateways intelligently coordinate local batteries to charge or heat pumps to buffer thermal energy. During the expensive evening spike, the fleet discharges power or throttles heavy charging loads. By coordinating demand in real time, we lower energy costs and open the door to integrating ever more clean energy into the grid. Doing this at a large scale demands absolute edge availability."

Reliable IoT Messaging Infrastructure

Heartbeat AI depends on a stable path between gateways and upstream cloud services. Internally, the IoT Platform team measures that as a single logical, bi-directional channel: from the gateway through an MQTT Adapter into a Cloud Native Message Broker.

Over the last six months, that Gateway-to-Cloud channel has stayed above 99.9% availability against the team’s service-level objectives. That number matters because it proves the control and telemetry layer behind the VPP is built to operate at scale. Yet a truly resilient VPP must pair this cloud availability with advanced safeguards at the edge level to ensure that, during short times of disconnects, the Gateways continue to function seamlessly, as well as minimize situations where devices become bricked, get stuck on a bad OS image, or are unable to receive updates.

From Cloud Abstractions to Edge Realities

For Simon, transitioning to the IoT Platform team required a major shift in mindset. "My background was entirely in Cloud Engineering," Simon reflects. Transitioning from cloud infrastructure to localized edge deployments required a radical change in engineering mindset. "In a standard cloud environment, a failing microservice is trivial, Kubernetes simply kills the container and schedules a replacement on identical, abstracted hardware inside a controlled data center. You are entirely shielded from physical constraints."

Residential garages and basements do not provide these abstractions. The 1KOMMA5° Edge services run on a System-on-Module (SoM), operating through conditions like unexpected power cuts or flaky networks.

"A bad deployment in the cloud takes seconds to roll back," Simon notes. "But a corrupted update at the edge can completely brick a device, destroying its economic model because it requires a technician to physically drive out and replace it". To safeguard device availability while allowing product teams to ship updates frequently, 1KOMMA5° implemented a three-layer decoupling pattern:

1. System Layer (Edge OS): stability and controlled change

The base operating system is a lean custom Linux distribution compiled via the Yocto Project framework. To prevent runtime configuration drift on thousands of hosts, and eliminate unexpected vulnerabilities, the base filesystem is entirely read-only. System updates utilize a fail-safe, dual-image A/B partition scheme managed through U-Boot.

If an OS update fails in the middle of the process, or post-reboot health checks don't pass, a hardware watchdog timer steps in, and automatically falls back to the previous known-good partition.

Because remote recovery and further rollouts depend on the device still being able to talk to the Mender server, the team treats Mender client health after an OS update as a gate before committing the new rootfs as the long-term boot target. If the device cannot complete the expected post-update checks for Mender health, the safe default is to not treat the rollout as complete. This allows the team to always retain a path to diagnose and fix gateways remotely.

Reboots for OS updates are acceptable here; avoiding unrecoverable states through A/B updates matters more than avoiding every restart.

2. Application Layer: deploy apps often, touch the OS rarely

Energy features and asset integrations change often: new inverters, revised drivers, improvements in the energy management algorithms, or changes in regulatory requirements. Therefore, the Heartbeat AI energy features run in lightweight Docker containers managed with Docker Compose on top of the Edge OS.

Container rollouts let the team ship application updates frequently with minimal disruption compared to full OS images – without the full-device reboot that an OS slot swap implies. This separation allows product velocity, while the underlying OS maintains a more conservative and stricter update schedule.

"By providing this framework to our development teams they get an environment that mirrors cloud-native serverless runtimes like Google Cloud Run," Simon notes. "Developers can safely push features, update specific services, and test new logic to a container runtime without the need to deal with Edge complexities and full system reboots."

3. Embedded Observability Layer

Operating a large-scale VPP requires deep field visibility. Therefore, the gateway runs the “Edge Observability Stack” streaming logs, metrics, and traces to the 1KOMMA5° observability system. There, they are aggregated and used to automatically alert on fleet health issues and hardware-oriented signals.

This allows the team to accurately spot early indicators of physical degradation, such as high eMMC storage wear, as well as functional regressions. By doing so, the team ensures that anomalies are detected and neutralized before they can trigger fleet instabilities.

Within this architecture, Mender serves a distinct role as the essential safety net for the device lifecycle. While precise field telemetry catches anomalies early, the decoupled OS layer including the Mender Client serves as the final fail-safe mechanism. If an unpredictable software regression breaks an optimization microservice, the Edge OS remains completely unimpacted, preserving secure, authenticated troubleshooting capabilities via Mender for instant point-fixes over-the-air or larger scale rollbacks and hot fixes.

Minimizing Deployment Risks

To advance Heartbeat AI's mission of integrating more renewables, 1KOMMA5° brings a diverse array of energy assets onto the platform. Real-world residential sites feature a fragmented matrix of different energy asset combinations, making their integration a demanding operational challenge: different PV inverters, batteries, heat pumps, and firmware quirks of the energy assets paint a heterogeneous picture.

Therefore, to contain the systemic impact of new updates across these varied configurations, the team leverages Mender to orchestrate a progressive delivery pipeline:

  1. Hardware-in-the-Loop (HIL) Testing: New release candidates are first validated in dedicated test environments against real physical assets, ensuring the business logic interacts safely with physical hardware before any code touches a customer home.
  2. Canary Deployments: Updates are then introduced to a controlled "Canary Group" – a representative fleet of internal devices running diverse hardware permutations to catch edge-case anomalies in real-world scenarios.
  3. Phased Production Rollouts: Software is deployed to the broader production fleet in gradual, automated percentage increments. This rollout is continuously guarded by system monitors capable of halting updates if any regressions are detected.

Each step maps back to the same idea: prove safety on small, representative surfaces, then widen.

The success of Heartbeat AI in mitigating the Duck Curve is fundamentally rooted in this IoT platform engineering philosophy. Through a three-layer architecture, A/B updates for the Edge OS, leveraging Mender as a foundational safety net, and strictly containing the systemic impact of new updates, 1KOMMA5° ensures the edge reliability that is required for VPP orchestration accelerating the shift to clean energy.