12-03, 14:00–14:30 (UTC), General Track
Non-Intrusive Load Monitoring (NILM) is a key technique in data-driven energy management and home automation, aimed at disaggregating energy consumption to identify active appliances in households and quantify their energy usage. This presentation:
- Provides an overview of NILM, highlighting its advantages and reviewing state-of-the-art deep learning algorithms developed for this purpose.
- Examines smart meters and IoT devices in energy systems, with a focus on the Chain2 protocol used in Italian energy systems. This event-based protocol generates low-volume data, enabling real-time energy monitoring and alerting.
- Presents examples of deep learning models trained on real-world IoT sensor data from energy meters, demonstrating their application in energy disaggregation.
This session offers an insightful overview of real-world deep learning applications in energy systems. While tailored for data scientists and data engineers interested in these fields, no prior knowledge is required. Join to explore how these technologies are driving energy optimization, cost reduction, and enhancing personal energy consumption awareness.
Non-Intrusive Load Monitoring (NILM) aims to track the energy consumption of individual household appliances, helping to increase awareness of energy savings, reduce waste, and improve smart home automation through integrated domotic systems.
Currently, scientific research on NILM mostly relies on synthetic datasets or small-scale real-world datasets from a few homes, often using energy clamps rather than actual data from energy meters.
This presentation explores these topics, with a focus on how we implemented a real-world NILM service for one of our clients using IoT devices that gather data in near-real time from smart meters. The presentation is structured as follows:
- The first part introduces the NILM problem, explaining its benefits for both consumers and energy providers, while also highlighting the challenges in this field. (Approx. 5 minutes)
- The second part discusses the protocols and standards implemented in Italy in response to the EU energy metering directive. Specifically, it highlights the Chain2 protocol, which enables IoT devices to access real-time energy measurements from smart meters, allowing for real-time monitoring of energy consumption. It also compares energy clamps with smart metering, emphasizing the advantages of smart meters. (Approx. 10 minutes)
- The third part focuses on state-of-the-art approaches for NILM and explains how our company developed a NILM service. It details strategies for effectively training deep learning models on real-world data to classify active appliances. (Approx. 10 minutes)
Audience:
This session is for anyone interested in deep learning, IoT, or energy management, whether you are a data scientist, data engineer, an energy expert, or simply curious about how these technologies work together.
Type and Tone:
The talk is informative and focuses on real-world examples, demonstrating practical applications of deep learning in energy systems.
Takeaways:
You will learn how deep learning is changing the way we manage energy, how IoT devices are used for real-time monitoring, and how these advancements can help save energy and reduce costs in everyday life. Additionally, you will become more aware of your own energy usage!
No previous knowledge expected
I am a Data Scientist who loves deep learning, MLOps, and working with data. I enjoy participating in conferences and competing in Machine Learning challenges to continually improve my skills. I also love traveling and visiting new places around the world.
I began my career as a Telecommunications Engineer with a background in Statistical Signal Processing, pretty tough stuff! After working for 3 years in university research and industry R&D, I joined AgileLab, an Italian consulting company with the mission of "elevating the data engineering game and empowering companies to shape their future around data."
I believe that successful data science projects should be built on solid software engineering and data engineering practices to ensure effectiveness and reliability.