A Smart Approach to Energy Disaggregation

John Marino, General Manager, LEM USA


A non-intrusive method for obtaining appliance-specific energy usage data at a building’s breaker level without multiple submeters.

Figure 1 – Traditional energy disaggregation monitoring requires several submeters; NIALM monitoring happens at the main breaker level.

In today’s world, careful and intelligent energy management is critical.Rising energy costs and environmental pressures are all driving corporations to examine their energy use more closely and, in turn, driving the development of advanced energy management systems.

Demand for compliance with green building rating systems (LEED, BREEM) has been joined by increasing legislative pressure to implement smart building energy monitoring: EU’s energy efficiency directive and various governments’ enforcement of the ISO50001 energy efficiency standards are two broad examples. California's Title 24 mandates that that electrical power distribution systems measure loads downstream of the service meter, such as tenant areas and lobbies. France’s green building law requires each residential dwelling to measure the energy consumption of heating, air conditioning, hot water, electrical circuits and other systems.

Predictably enough, the first step in implementing any energy conservation measures (ECMs) is measurement. Industrial companies, factories and multi-tenant buildings need to understand how their properties are using electricity in order to develop strategies aimed at increasing energy efficiency. To that end, they are increasingly installing submetering equipment to collect real-time, in-house energy usage data to accurately calculate expenses and allocate costs.

Real-time energy information and energy disaggregation-based solutions can influence consumer behavior to increase savings and drive engagement, all at a lower cost than traditional submetering methods. Energy disaggregation technology can also be used to perform remote energy audits, measure and validate utility demand-response programs, and resolve high bill disputes between the utility and the building owner.

Smart Building Energy Disaggregation

Using energy disaggregation to acquire detailed energy usage is not new. But until recently, it has been complicated. Traditionally, energy monitoring required installing several multi or single point submeters in each supply closet, each capable of monitoring 1 to 3 phase circuits (see Figure 1). Depending on factory of building layout, electrical distribution may be spread throughout the building, requiring several submeters mounted in close proximity. This method is costly, intrusive and requires significant installation and maintenance efforts.

A more innovative approach to obtaining appliance-specific data is through the Non-Intrusive Appliance Load Monitoring (NIALM) method. A NIALM submeter undertakes the real-time power consumption breakdown and analysis at the main breaker level, using a statistical approach to processing that single power measurement to extract equipment and appliance-level data without the need for plug-level sensors (Figure 2). It then transmits energy consumption data to a gateway that sends the readings to cloud-based storage for the building manager to access and analyze.

Click image to enlarge

Figure 2 – A NIALM submeter uses a statistical approach to extract equipment and appliance-level data from a single power measurement.

A NIALM submeter and gateway, paired with high-accuracy, low phase shift current transformers, can collect an entire building’s consumption data. By using a split-core CT with clip-on mounting, there is also no need to interrupt the power for installation. The CT sends a signal to the NIALM submeter, which carries out real-time power consumption breakdown and transmits that data to a gateway. The data is then uploaded by the gateway for analyzing with energy management software.

New Ferrite Compounds

NIALM monitoring is not a new concept.  However, for the last two decades there has been an explosion in NIALM patents – the number of patents tripled between 2006 and 2011 alone.   So, what has changed? Until recently, current transformers with both high accuracy and low phase-shift have been too costly to implement the technology on a large scale. Recent developments have greatly improved the ability of new ferrite compounds to perform at these frequencies and brought the advantages ferrite to a wide range of power monitoring applications.

Although ferrite compounds have been used in CT sensors for years, their poor performance in terms of saturation level and magnetic permeability made them unusable at frequencies as low as 50/60Hz.

The new types of ferrite have significantly improved permeability and can be implemented in 50/60 Hz current transformers as a substitute for FeSi or FeNi cores, despite the low magnetic saturation level. Split-core current transformers such as LEM’s ATO series use these new ferrite types to accurately measure AC signals in an extended frequency range that includes 50/60 Hz. In addition to taking advantage ferrite’s intrinsic ability to provide high accuracy and excellent linearity even at very low current levels, they also feature particularly low phase-shift between input and output currents, which is essential for accurate measurements of true active power or energy. The dense core allows air gaps to be minimized and is virtually immune to ageing and temperature changes, in contrast to materials like FeSi or FeNi.

Class Accuracy

The increased sensor accuracy provided by these newer ferrite compounds is giving NIALM applications a boost. Current sensors now fall under the IEC 61869 standards, with IEC 61869 having essentially replaced IEC 60044 standards over the past few years. Although the two standards are almost identical, IEC 61869 differentiates between Class 0.5 accuracy and Class 0.5s accuracy.

IEC 61869 class accuracy requires two criteria for compliance: ratio error and phase displacement. Accuracy tests require not only measurement accuracy of +/-1% of the primary rating (IPr), but also at 5%, 20% and 120% of the rated primary current. Moreover, phase displacement at the same rated primary currents has to be respected. Phase shift becomes incredibly important at lower power factors, as even small errors due to phase shift can render the data useless.  

Figure 3 gives an example of IEC 61869 Class 1.0 accuracy using the ATO CT. For a 75A waveform nominal current, 1% and 3% accuracy are required when measuring respectively at 120% and 5% of IPr.

Click image to enlarge

Figure 3 – For class accuracy under IEC 61869 standards, two criteria must be met: ratio error and phase displacement.


Today the overall signal-to-noise percentage of correctly reconfigured appliances with NIALM algorithm based submetering solutions is around 80-90% and keeps improving thanks to new ferrite compounds and more accurate current sensors.

The more sophisticated the consumption metrics available, the more plentiful the benefits. More and more, current sensors are being used to ensure the reliable integration of distributed renewable energy, energy storage, production and consumption. With this will come greater monitoring at the control room level in order to track real-time data and optimize automation. In addition, energy conservation measures such as home energy management (HEM), battery monitoring systems (BMS) and electric vehicle stations will continue to expand the need for reliable, cost-effective energy disaggregation at the building level.