Boost home heating efficiency: Connecting millions of smart radiator thermostats via AWS IoT Core for LoRaWAN

The EU’s Green Deal sets clear targets for CO2 reduction: the EU aims to become climate neutral by 2050. In 2019, the EU emitted 4,065,462,000 tonnes* of greenhouse gases. The building sector accounts for a significant share of around one-third of all primary energy consumption, 80% of which is attributable to heating water. The efficient, energy-saving operation of heating systems in existing buildings is essential.

Europe is a continent built over centuries. Its existing buildings are culturally and regionally diverse and have a significant impact on the success of climate protection in the building sector. These existing buildings consume the most energy per square meter. New energy-insulated buildings represent only a fraction of our building stock. Smart technology solutions are the key to transforming this historic and diverse building stock at the required pace.

The noventic solution is the answer to this specific challenge: how do we achieve rapid savings in the central heating of existing, averagely insulated, multi-party buildings?

Solving the heating energy efficiency problem

As a group of companies operating across Europe, noventic’s mission is to bring greater clarity, efficiency, and simplicity to energy consumption in buildings and homes for everyone. This is accomplished with the help of digitalization, appropriate hardware and wireless infrastructure, as well as housing management services. beyonnex.io, noventic’s subsidiary with a strong tech-DNA, is implementing these solutions.

noventic has implemented the market’s first smart thermostat solution for apartment buildings. As a retrofit solution, the smart thermostats can be used with any standard heating supply in apartment blocks, which serves a wide range of heating infrastructures. On average, the energy savings achieved are up to 15%.

The smart thermostats enable hydraulic balancing: the first continuous algorithmic optimization to improve heating efficiency in apartment buildings. This makes sure that hot water is evenly distributed to all radiators in a building. Smart thermostats continuously adjust the valves on each radiator, making sure that every room receives the right amount of heat. This improves comfort and efficiency while reducing energy costs and usage.

A smartphone app controls the room temperature, developed specifically for the cross-section of society to reduce the entry barrier as much as possible. As the technological backbone, a private LoRaWAN wireless network connects the smart thermostats throughout the building to the cloud.

With increasing demand (including legal requirements) to improve energy efficiency, the product is an optimal fit for the housing industry on a large scale. The noventic group, with its subsidiaries, offers a full-service solution to the industry that includes assessment, installation, operation, and resident onboarding for this product.

Technical overview

The technical solution includes a range of components. Unique is the custom-designed Smart Radiator Thermostats (SRTs) with LoRaWAN connectivity, which is optimized for the housing industry. It is comprised of intuitive handling, a robust design, and up to six years of battery runtime. The complete solution also includes LoRaWAN Gateways, backend software components including a digital twin solution, as well as mobile apps for both residents and professional users (used during the installation process of the Gateways and SRTs).

beyonnex.io uses a range of AWS services to implement the solution, including Amazon Elastic Kubernetes Service (Amazon EKS), various databases, Amazon Cognito, among others, as shown in the following figure. Developed as a distributed system by a dozen teams, the majority of the communication is handled over Kafka.

Of particular importance is the AWS IoT Core for LoRaWAN, which provides the required LoRaWAN network components for device (SRT and Gateway) provisioning, as well as communication and Firmware Update Over The Air (FUOTA) as a serverless solution.

AWS IoT Core for LoRaWAN is a fully managed LoRaWAN Network Server (LNS) that allows customers to connect LoRaWAN devices and gateways to the AWS Cloud. This enables low-power, long-range connectivity without the need to develop, maintain, or operate their own server.

The reason for using LoRaWAN over alternative technologies was the cost and energy efficiency, as well as the widely available hardware solutions. Neither mobility nor higher bandwidth or lower latency, which other solutions can offer, are benefits when implementing the smart thermostat solution.

AWS IoT Core for LoRaWAN handles all device communication and management on the LoRaWAN level.

In the “uplink” direction, the devices send sensor updates, detect opened windows, and deliver heartbeat messages. AWS IoT Core for LoRaWAN is configured to directly publish those uplink messages to an Apache Kafka broker. The payload sent by the SRT devices is in the Protobuf format, and the smart heating backend deserializes this accordingly, updating the digital twin of the devices based on the received device status.

The “downlink” is used to send commands or configurations to the SRTs. For example, when the resident changes the room’s target temperature or configures the SRT heating schedules in the mobile app. Then, the smart heating backend converts the command to the Protobuf format and directly uses the AWS IoT SDK to send the command to the device. AWS IoT Core enqueues the commands for a device until the LoRaWAN device opens its next downlink receive window (for example after an empty uplink heartbeat message), and then sends down the commands, emptying the command queue.

Figure 1: Overview of the beyonnex.io smart heating architecture

Digital twin

Device data (of SRTs and Gateways) is mirrored in the digital twin IoT middleware Eclipse Ditto, which is operated by beyonnex.io in Amazon EKS, using MongoDB Atlas as persistent storage.

Ditto not only provides APIs for device interaction (providing access to the most recent and historical state and routing commands to devices), but also serves APIs for rooms, apartments, and buildings. By putting devices in relation to those entities, it is possible, for example, to query Ditto for the firmware version of all SRTs within a building.

The digital twins of devices, rooms, etc., provide a flexible abstraction for the upstream systems (resident app, on-site installation and maintenance app, customer portal, and internal operations dashboards), including the required granular authorization concepts (for example, by granting residents control over devices and rooms only within their apartment, or by allowing a building manager to control the room temperature only in vacant apartments).

In its role as an IoT middleware, Ditto also consumes commands (for example to change the state of a twin) and emits events (for example when the state of a twin was changed) through Kafka. It is also configured to forward commands sent to the devices cluster-internally to the smart heating (SRT control) backend. This translates those commands to the Protobuf format and sends them to AWS IoT Core for LoRaWAN.

By using an open standard (the W3C Web of Things (WoT) Thing Description 1.1), the devices and other entities managed within beyonnex’s Ditto are fully described by WoT “Thing Models”. These models describe the device’s capabilities, for example which state data (properties) are provided in which data type, unit, and restrictions at the Ditto APIs. They also describe which commands (actions) can be sent to the devices and other entities, and which events they emit. The uplink consumers in beyonnex’s IoT landscape benefit from these Thing Models, for example by being able to generate code and HTTP API specifications from them.

Scaling and FUOTA

After a successful Proof of Concept (PoC) for the backend technology, a soft start of the product was conducted over the winter of 2023/2024. Approximately 4,500 SRTs and 100 Gateways serving over 900 apartments have been deployed in the field. The goal is to reach almost 50x scaling in the coming winter, with the potential of reaching up to 1 million SRTs by the end of 2025.

The projected scalability was taken into account when designing the solution from the beginning. The seasonal workload (heating control only required in winter) is a textbook example for using flexible cloud-based infrastructure to elastically scale capacity when needed.

The development of new features, as well as improvements to the overall solution, often needs changes to the SRTs’ firmware. This could include things such as energy optimization of motor control or radio communication, or updates to the user interface (UI) of the SRT. Due to the conflict between the required long battery life (six years under normal use) and the energy consumption of firmware updates over the air, this leads to various firmware versions being deployed in the field over time.

Therefore, handling FUOTA for the SRTs is a key aspect in designing and operating the overall system. AWS IoT Core for LoRaWAN provides the necessary functionality to flexibly manage the FUOTA rollouts across selected groups of devices.

Adding reliability on top of LoRaWAN

Due to the unreliable nature of LoRaWAN, any messages exchanged with the device (in both uplink and downlink directions) may be lost. To make sure that important commands (for example changing the room target temperature or the schedule of temperature adjustments) reliably reach the SRT devices, an application-level mechanism is needed.

The beyonnex.io smart heating backend uses Ditto’s “desired properties” to reflect a pending (not yet confirmed) command for the device. The following figure shows the last reported/confirmed state of a device “feature” (here the child-lock functionality) on the left side, and the desired (pending) state on the right side.

The SRT device confirms each command once it’s received through a so-called “mirror uplink.” Upon successful confirmation, the smart heating backend clears the desired property and applies the property to the digital twin’s state. If no confirmation is received within a certain time interval, then the backend retries sending the command until it is confirmed, providing “eventual consistency” of delivering commands to the SRT device.

Fig 2: Snapshot of the digital twin while the child-lock disabling command was not yet confirmed by the SRT device

This approach also provides transparency for uplink services with changes that aren’t yet applied to devices, but will be applied. This is so that for example the resident app can already show the resident that the wish for an increased room temperature is processed.

Conclusion

noventic’s innovative use of AWS IoT Core for LoRaWAN demonstrates the immense potential of IoT in transforming energy efficiency in apartment buildings. By connecting millions of SRTs to the cloud, noventic uses advanced analytics and machine learning (ML) to optimize heating systems, resulting in significant cost savings and a smaller carbon footprint. This case study highlights the scalability, security, and efficiency of AWS IoT Core for LoRaWAN in creating smart, sustainable solutions.

*https://www.europarl.europa.eu/topics/de/article/20180301STO98928/treibhausgasemissionen-nach-landern-und-sektoren-infografik