An innovative meter system which detects individual devices out of the overall power consumption –  this is the challenging task that the affiliated partners Discovergy GmH, EasyMeter GmbH, Green Pocket GmbH and RWE GBS GmbH under the leadership of the  Fraunhofer Institute for Microelectronic Circuits and Systems IMS dedicated themselves to. During the two-year research project NILM (Nonintrusive Load Monitoring), which started in October of last year, machine learning techniques / data mining methods are extracted out of high-frequency measuring data. The system is significantly cheaper than submetering and is being tested with test customers from industry, trade and service (IGHD) which is the clientele that it is specifically designed for.

Having breakfast on a Sunday morning with fresh bread rolls and reading the newspaper. From production to the point of sale of newspapers, baked goods or even the table – all of the machines, like cutting systems, ovens and cooling systems consume a lot of power. This can be seen at the example of a bakery: large-sized mixing and kneading machines, hydraulic systems and ovens are in use to bake fresh bread rolls from dough.

According to the energy institute for economy, there is a savings potential of up to 30 percent in bakeries, especially in the area of ovens and waste heat. According to a meta study regarding 36 individual studies from North America and Europe out of the whole industry, business and trade sector, long-term savings of more than 12 percent are expected, provided that the power consumption of individual devices are known and analyzed. The study: “Energy efficiency: Potentials, economic effects and innovative fields of action and support for the national climate protection initiative” expects the greatest saving potentials in Germany in the use of energy-efficient pumps. (5 billion kWh), sustainable lighting systems (12 billion kWh), effective ventilation (4 billion kWh) and compressed air systems (5 billion kWh) as well as efficient electric motors and other engine systems (52 billion kWh).

In the course of the research project “NILM – Nonintrusive Load Monitoring”, supported by the Federal Ministry for Economic Affairs and Energy (BMWi), a smart meter that breaks down the overall power consumption into the distinct devices is developed. The partners of this joint project under the leadership of the Fraunhofer Institute for Microelectronic Circuits and Systems IMS develop the corresponding hardware as well as the detection algorithms. The single device identification is then integrated into an energy management software.

Fingerprint in the power supply

The central idea of this project is that every machine or every device influences the power and voltage in a characteristic way and therefore leaves a type of “fingerprint” inside the power grid. This fingerprint is captured as an aggregated overall power consumption by the measurement technology, the smart meters. Using data mining, patterns in the power consumption are detected and can be assigned to devices. In a laboratory test system devices are recognized easily when the allocation of the device-specific power consumption has been activated by turning it on, or a constant active power has been used from the power grid (permanent consumer). In the current research project NILM machine learning methods for pattern recognition are being used which extract and analyze the consumption pattern of devices in the IGHD sector out of the overall power consumption automatically via different electrical parameters. Through this combination of pattern recognition algorithms and machine learning methods NILM technology improves and simplifies the process of disaggregation of power consumption. The application of several sub meters for measuring current flows is therefore unnecessary – one meter is enough.

Use and application errors

In comparison to using several meters, also called submetering, the NILM system identifies the devices with the highest power consumption using only one meter. Not only does that save the costs for expensive measurement hardware, installation and maintenance, but it also saves electricity costs. When the power guzzlers are identified, saving potentials can be derived. Using the high-frequency measured data, it is easy to detect which operating times the devices have and how peak loads can be avoided by delayed activating/deactivating of devices (peak shaving).

After the disaggregation the measurement data is provided in real time. Due to the constant monitoring (condition monitoring) of the energetic device condition, malfunctions of a device can be detected prematurely and therefore an impending production stop can be prevented by taking timely mannered countermeasures.

Potential malfunctions are displayed by the specific device consumption pattern. If the electric parameters of the high-frequency measured IST consumption pattern deviate from the typical power consumption pattern, it is possible that a plant section broke down or is defective. When the responsible people are contacted automatically, countermeasures can be taken and therefore production losses can be prevented. The breakdown of the power consumption also allows for the distribution of energy costs according to the source. That means that every device can be assigned its exact energy costs that were used for the process and the operation of the device, respectively. Based on this information, the analysis can improve the efficiency and the processes of specific devices. Accordingly, entire production processes can be optimized, as correlations between devices as well as changes in process steps can be displayed more efficiently than in the submetering procedure.

The project

During the two-year research project, which started in October of last year, the project partners are developing a NILM meter system under the leadership of Fraunhofer IMS. The system consists of an extensible smart meter, a gateway and software for consumption analysis. The high-end NILM system, which is being tested in corporations of the business sector IGHD until the end of the year, has a sampling rate of up to 1 megasamples per second (MS/s). The sampling rate is a measurement for how much information is available in a particular signal. During this test phase electrical parameters are being extracted out of high-frequency measuring data in the test corporations. These parameters are then examined for their effectiveness on disaggregation using different algorithms from the area of machine learning. After the extensive test of the high-end systems, the obtained knowledge is transferred into the development of a more cost-efficient NILM prototype. The project partners complement each other perfectly and are working together on the design of the NILM system:

• Fraunhofer IMS (direction): Research, development and pilot manufacturing of microelectronic solutions
• EasyMeter GmbH: Development and production of sustainable, modular meter solutions
• Discovergy GmbH: Supplier of intelligent smart meter solutions and national meter operators
• GreenPocket GmbH: Software specialist for IoT and smart energy solutions
• RWE GBS GmbH: Efficient, climate-friendly and intelligent power supply

The further developed, NILM-capable smart meters with high sampling rates are being developed by EasyMeter GmbH. The gateway and processing servers are being developed by Discovergy GmbH. GreenPocket GmbH is supporting the algorithm development, developing the consumption data analysis and integrating the NILM technology into the existing energy management software for commercial customers. The test customers in the commercial and industrial sector are provided by RWE GBS GmbH.

This innovative NILM system, which is being developed in cooperation with all project participants, offers additional benefits for the IGHD sector and has the potential to inspire the measurement technology and smart meter market over the long term.

About Fraunhofer IMS

For 30 years researchers at Fraunhofer IMS in Duisburg have been developing microelectronic circuits, electronic systems, microsystems and sensors. Due to the extensive know-how, the access to technology and the high-quality development performance, the institute is a globally recognized partner in the industry. In eight business fields Fraunhofer IMS dedicates itself to applied research, advance development for products and applications. Stable, efficient and marketable technologies and processes, which are used in many sectors, are the main focus of the commissions.

About Green Pocket

GreenPocket is a specialist for software applications in the Internet of Things and energy management solutions in the smart energy area. With over 100 successful projects for more than 50 national and international customers, GreenPocket is one of the leading suppliers in the smart energy market.

More information on www.NILM.info