Nordic-powered smart thermostat employs machine learning for intelligent home automation

eCozy 2.0 uses Nordic nRF5340 SoC for Matter-over-Thread, Zigbee, and Bluetooth LE connectivity between device, gateway, and user’s smartphone

Germany-based eCozy has released a machine learning (ML)-powered smart thermostat solution for water heating radiators in the smart home automation market. ‘eCozy 2.0’ is designed to provide intelligent heating control and enable remote energy consumption monitoring, helping users save up to 30 percent on annual heating costs, according to the company.

The eCozy 2.0 platform comprises a smart thermostat device powered by the Nordic Semiconductor nRF5340 SoC, alongside a central unit (connected to the home’s Internet router via Wi-Fi) that performs as a wireless gateway to the Cloud and can be used as a Matter controller, and an associated app for remote system management.

The smart thermostat physically replaces the existing thermostatic head on a water heating radiator and contains a stepper motor with plunger to open and close the radiator valve. In a reflection of the company's design-oriented approach, the eCozy products have been awarded the prestigious RedDot Design Award twice. The device features two integrated high-precision temperature sensors, optional external temperature sensor (via Matter-over-Thread), high-precision humidity sensor, and high-level accuracy microphone.

Matter-over-Thread and Bluetooth LE connectivity

eCozy 2.0 employs the Nordic nRF5340 SoC's advanced multiprotocol radio for wireless communication via Matter-over-Thread and Bluetooth® LE. The hardware has also been architected to support Zigbee in future firmware releases, ensuring future-proof interoperability. This multiprotocol connectivity enables the smart thermostat to connect to the gateway (eCozy Central Unit and other gateways based on Matter-over-Thread) in the smart home, which in turn securely relays the sensor data to eCozy’s Cloud servers, and to the user’s smartphone from where the associated ‘eCozy’ app can be used to view the information and remotely control the device. 

From the app, users can configure the system, adjust settings, create detailed heating schedules for each room, monitor current temperatures, and view energy consumption reports. The solution is also compatible with major voice assistant platforms like Amazon Alexa, Apple HomeKit, and Google Assistant, allowing users to control and monitor the temperature in specific rooms with simple voice commands.

“The primary benefit of eCozy is the significant energy and cost savings – by heating only when and where it is needed the system can reduce heating bills by up to 30 percent, and elimate the common problem of heating empty rooms or overheating occupied ones,” explains Andre Kholodov, CEO of eCozy GmbH.

Quote
Nordic Semiconductor’s integrated approach—including powerful development tools and reference designs—can simplify the development supply chain and reduce integration risks
Andre Kholodov, CEO of eCozy GmbH

Nordic nRF5340 SoC powers on-device ML 

A number of eCozy 2.0’s features and functionalities—including anomaly detection and adaptive heating schedules—are enabled by a powerful combination of on-device ML capabilities, powered by the Nordic nRF5340 SoC, and advanced AI algorithms running on the eCozy Cloud backend. For example, the eCozy system could detect unusual heating patterns or potential issues (open window, smoke alarm, fire, etc.) and alert the user, or learn user behavior patterns (such as when they are typically home, away, or asleep) and automatically optimize heating schedules for improved comfort and energy savings beyond pre-set rules. The on-device ML processing also enables the high-accuracy microphone to perform advanced acoustic event recognition, such as identifying the distinct sound of a smoke alarm and triggering an immediate alert to the user's smartphone.

The nRF5340 SoC features dual Arm® Cortex™-M33 processors; a high-performance application processor capable of DSP and Floating Point (FP) for supervising the device’s integrated sensors, as well as a fully programmable, ultra-low power network processor providing the Matter, Zigbee, and Bluetooth LE connectivity. The hardware has been designed to support future integration of an energy harvesting module for long-lasting, eco-friendly performance with minimal maintenance (available now as minimum viable product).

The eCozy 2.0 smart thermostat employs four AA alkaline batteries, achieving extended battery life of up to two years between recharge, thanks in part to the ultra-low power operation of the nRF5340 SoC. The smart thermostat is equipped with a colored LED display and touch sensor buttons for on-device control. For added flexibility, the device is also equipped with a USB-C port, allowing for a continuous power supply.

“In addition to the key hardware and software features of the nRF5340 SoC, Nordic Semiconductor’s integrated approach—including powerful development tools and reference designs—can simplify the development supply chain and reduce integration risks compared to sourcing components and software from multiple vendors,” says Kholodov.

“For example, the comprehensive Nordic nRF Connect SDK [Software Development Kit] provides a ready-made, integrated environment with a RTOS [Zephyr], protocol stacks [Matter, Zigbee, Bluetooth LE], hardware abstraction layers, drivers, libraries, and application samples. This saves significant development time compared to building these components from scratch.”