**6. Conclusions**

14 Will-be-set-by-IN-TECH

Normalized current

Normalized voltage

(b)

Microwave

Normalized voltage

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

However, higher sampling rate is desirable if transients will be also analyzed.

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.

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

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

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

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

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

(a)

Air conditioner

Normalized current

found or the generation limit is reached.

Bayesian Network (DBN) is applicable.

non-sinusoid signatures.

known.

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 that lead to effective energy savings.

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

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 to promote permanent energy savings.
