**5. Centralized measurement methods**

With the main objective of minimizing the number of sensors needed to monitor all appliances and also reducing the complexity of the installation service, many researchers are proposing centralized measurement approaches. There are several proposals of different techniques to identify the appliances connected in the same circuit.

In [22] the authors presented the design and implementation of a wireless monitoring system for residential spaces, based on a a multi-layer decision architecture called TinyEARS. This system was able to recognize the appliances by deploying one acoustic sensor node per room, that will identify the appliances' acoustic signatures. Combining this information with the data acquired by a real-time power meter installed at the main electric panel and with relatively simple processing algorithms the system can recognize the appliances with an overall success rate of 94%.

Another technique widely adopted is the identification based on load signatures. Load Signature is an electrical expression that an appliance distinctly possesses regarding its electrical consumption behavior. It can be measured in various forms. From power consumption levels to waveforms of electrical quantities such as voltage and current. Almost every electrical measurement can be treated as a load signature. It can be represented in the frequency domain [6], in the time domain [11] and can also be represented mathematically in terms of wavelets, eigenvalues, or components of the Singular Value Decomposition (SVD) [12].

Methodologies which use signal processing techniques and estimation algorithms for signal load recognition based on load signatures allows the use of a single intelligent device in only one point of the installation (in the electric panel). The detailed energy breakdown of the whole installation is then calculated by sophisticated algorithms.

One of the earliest works (1980's) in nonintrusive monitoring was developed at MIT and had its origins in load monitoring for residential buildings [9]. In the developed technique the operating schedules of individual loads are determined by identifying times at which electrical power measurements change from one steady-state value to another. These steady-state changes, known as events, correspond to the load either being turning on (or turning off), and can be characterized by the magnitude and sign in real and reactive power values. Figure 9 shows the power consumption of a refrigerator and a microwave oven, where two different-sized step changes are clearly present, providing characteristic signatures of the refrigerator and the microwave oven. Knowing the time of each on and off event, it is possible to determine the total energy consumption of the refrigerator and the microwave oven.

12 Will-be-set-by-IN-TECH

In the case presented in table 3, the electricity bill was monitored during the next six months after the results were presented. The family reported that, after the report was presented, all three PCs were turned-off during the night and the timer of the bedroom TV was set to turn-off the TV set at midnight.In the case of the The monthly savings in the electricity bills during these six months period were over 20%, when compared to the same month of the previous year. This result is a clear evidence that if a customer is informed of a habit that is wasting energy and this habit can be easily changed, a lot of energy can be saved and and this

The measurement system was designed to get the information from every appliance in the house in real time, but the customers reported that they were interested only in the two weeks final result. Thus, the ZigBee network could be removed from the smart meters and a simple peer-to-peer communication (from each smart meter to the DPU) could be used, reducing the

With the main objective of minimizing the number of sensors needed to monitor all appliances and also reducing the complexity of the installation service, many researchers are proposing centralized measurement approaches. There are several proposals of different techniques to

In [22] the authors presented the design and implementation of a wireless monitoring system for residential spaces, based on a a multi-layer decision architecture called TinyEARS. This system was able to recognize the appliances by deploying one acoustic sensor node per room, that will identify the appliances' acoustic signatures. Combining this information with the data acquired by a real-time power meter installed at the main electric panel and with relatively simple processing algorithms the system can recognize the appliances with an

Another technique widely adopted is the identification based on load signatures. Load Signature is an electrical expression that an appliance distinctly possesses regarding its electrical consumption behavior. It can be measured in various forms. From power consumption levels to waveforms of electrical quantities such as voltage and current. Almost every electrical measurement can be treated as a load signature. It can be represented in the frequency domain [6], in the time domain [11] and can also be represented mathematically in terms of wavelets, eigenvalues, or components of the Singular Value Decomposition (SVD)

Methodologies which use signal processing techniques and estimation algorithms for signal load recognition based on load signatures allows the use of a single intelligent device in only one point of the installation (in the electric panel). The detailed energy breakdown of the

One of the earliest works (1980's) in nonintrusive monitoring was developed at MIT and had its origins in load monitoring for residential buildings [9]. In the developed technique the operating schedules of individual loads are determined by identifying times at which electrical power measurements change from one steady-state value to another. These steady-state changes, known as events, correspond to the load either being turning on (or turning off), and can be characterized by the magnitude and sign in real and reactive power values. Figure 9 shows the power consumption of a refrigerator and a microwave oven, where

whole installation is then calculated by sophisticated algorithms.

power consumption of the measurement system and increasing the batteries' life.

savings tend to be permanent.

overall success rate of 94%.

[12].

**5. Centralized measurement methods**

identify the appliances connected in the same circuit.

**Figure 9.** Characteristic signatures of a refrigerator and a microwave oven sensed on the same circuit

In [12] the authors highlighted the fact that the complex electrical loads of today have signatures that vary with time, depending on their state and mode of use and that common appliances can have non-linear load characteristics. They propose a conceptual modeling to characterize an appliance based on three sets of signatures that are extracted from the appliance: steady state, transient state and operational pattern and therefore construct a taxonomy for the appliances.

A method to construct taxonomy of electrical appliances based on load signatures is presented in [10]. In this work the authors suggests a 2-dimensional form of load signature denominated voltage-current (V-I) trajectory to characterize typical household appliances. The V-I trajectory load signatures consists in acquiring the steady-state voltage and current in one-cycle long, normalize them to eliminate the effect of the current magnitude in the size of V-I trajectory, and then plot the V-I trajectory. After creating the trajectories for the appliances, the shapes of the trajectories of the appliances can be analyzed.

The proposed methodology for constructing the load taxonomy is summarized as: (1) the voltage and current waveforms of the household appliances are measured; (2) load signatures in the form of V-I trajectory are constructed; (3) shape features are extracted from the V-I trajectories; (4) hierarchical clustering method is applied to cluster the appliances; (5) the load taxonomy is constructed according to the clustering results.

In [13] the authors proposed a methodology of using load signatures and Genetic Algorithms (GA) to identify electrical appliances from a composite load signal. They introduced a classification method to group the appliances and how to disaggregate the composite load signals by a GA identification process from a generated random combinations of load signatures from the groups of appliances.

The methodology consists of defining a signature for each appliance by averaging 50 consecutive one-cycle steady-state current waveforms. Then the current waveforms are grouped by the ratio of their fundamental (50Hz) component verses their Root-Mean-Square (RMS) total, after a Fast Fourier Transform (FFT) calculation. That means that the higher the

circuit breaker. The technique under development presents a methodology to disaggregate the power consumption information of each appliance based on synchrophasors as load

Energy Measurement Techniques for Energy Effi ciency Programs 207

A compilation of several state-of-the-art methods that can be used to obtain detailed information about the energy consumed by the appliances in a household has been presented. The decentralized systems, where it is necessary to install one meter per home appliance, are more suitable to be employed in energy efficiency programs where the objective is to present a report of the breakdown of the electricity bill to the family. It has been observed that, based on the reports, many customers implement important (and sometimes simple) changes in habits

When it is desired to permanently monitor the household appliances in real time, for example to control the demand in small business installations or to shift the use of energy from peak periods to low-cost tariffs non-peak periods, usually the centralized systems are more efficient. However, these systems are not yet capable of performing a complete identification of all the

Advanced metering initiatives are fundamental for providing accurate information and understanding on how and where the energy is being used in a household, and also to evaluate the actual results of implemented energy efficiency programs. Educating people by showing how they can make a better use of the electrical energy seems to be the key element

*Department of Electronics and Microelectronics. School of Electrical and Computer Engineering of*

[1] *AD71056 Data Sheet - Energy Metering IC with Integrated Oscillator and Reverse Polarity*

[2] *ADE7953 Data Sheet - Single Phase, Multifunction Metering IC with Neutral Current*

[4] Darby, S. [2006]. *The effectiveness of feedback on energy consumption. A review for DEFRA of the literature on metering, billing and direct displays*, Environmental Change Institute,

[5] DEMIC/FEEC/UNICAMP [2012]. Monitoring power consumpion of in-house appliances by measuring the distribution board in circuit-level. Pre-proposal for a

appliances installed, and further investigations are required in these type of systems.

signatures.

**6. Conclusions**

that lead to effective energy savings.

to promote permanent energy savings.

This research is partially supported by CAPES.

*University of Campinas. Campinas, SP - Brazil*

*Indication* [2006]. Analog Devices.

University of Oxford.

*Measurement* [2011]. Analog Devices.

[3] BP [2009]. Statistical review of world energy 2009.

corporate-sponsored research agreement.

Luís F. C. Duarte, Elnatan C. Ferreira and José A. Siqueira Dias

**7. Acknowledgements**

**Author details**

**8. References**

**Figure 10.** Examples of V-I trajectories of (a) an air conditioner and (b) a microwave oven

ratio, the more sinusoidal shape of the signature is. They considered a sampling rate of 200 points per cycle (50Hz), which is sufficient for steady-state assessment in the time-domain. However, higher sampling rate is desirable if transients will be also analyzed.

The appliances identification can be done by making use of Genetic Algorithms, which are stochastic global search methods based on the principle of the fittest survives. A fitness function, which calculates the least sum squared error between the proposed aggregated signal and the measured signal, is firstly defined. A group of potential solutions are also defined as the initial population. Different variables/attributes, termed the genes, which would affect the fitness function, are allowed to cross-over and mutate to form new generation of potential solutions. In each generation, those matches having the highest fitness value would be retained for further reproduction. The process repeated itself until the best fit is found or the generation limit is reached.

The identification accuracy of the Genetic Algorithm technique is very high for a small number of appliances, but it decreases as the number of aggregated appliances increases. Also, the identification accuracy for sinusoid and quasi-sinusoid waveforms is lower than those of non-sinusoid signatures.

A different approach based on centralized measurement is presented in [14]. The proposed methodology detects the state changes of appliances and acquire energy information simultaneously. The appliance identification is performed using power meters installed in the circuit-level of the electrical panel, and measures the total electrical consumption every 5 seconds, assuming that all appliances and the states of theses appliances in the circuit are known.

In addition, user behavior is taken into account, as the algorithms assume that there are some patterns for using appliances. For example, when using the computer, first the user may switch on the light in the room, then start up the power of the computer and finally turn on the monitor. If the user has a regular lifestyle, the pattern is likely to be regular. Based on that assumption, the temporal character is taken into consideration, therefore, the Dynamic Bayesian Network (DBN) is applicable.

A recent proposal [5] proposes the use of several energy meters ICs, one for each circuit breaker in the electrical panel. This approach aims to minimize the number of appliances sensed in the same circuit breaker. If equal loads, which would present almost perfectly matched load signatures (for example, two air conditioning systems) are connected to different circuit breakers, the identification will be very simple, since it is performed per circuit breaker. The technique under development presents a methodology to disaggregate the power consumption information of each appliance based on synchrophasors as load signatures.
