APPARATUS AND METHOD FOR CLASSIFYING HOME APPLIANCES BASED ON POWER CONSUMPTION USING DEEP LEARNING

• US 20170116511A1
• Filed: 10/27/2016
• Published: 04/27/2017
• Est. Priority Date: 10/27/2015
• Status: Active Grant

• Current Assignee: Pusan National University
• Original Assignee: Pusan National University
• Inventors: KIM, Howon, KIM, Jihyun
• Granted Patent: US 10,572,797 B2
• Time in Patent Office: Days
• CPC Class Codes: G06N 3/044 Recurrent networks, e.g. Hopfield networks
• CPC Class Codes: H04L 12/2825 Reporting to a device located outside the home and the home network access arrangements H04L12/2856; for remote control or remote monitoring of applications H04L67/025; telephonic communication systems adapted for combination with telemetering systems H04M11/002

Abstract

Provided are an apparatus and method for classifying home appliances based on power consumption using deep learning, which can efficiently classify home appliances in use by applying deep learning and analyzing power data collected from a house. The apparatus includes a home appliance classification model creation module configured to encode power consumption data collected from a house to learn a home appliance classification model and create an RNN-based home appliance classification model and a home appliance classification module configured to collect and encode data on power consumption currently in use and classify home appliances using the home appliance classification model created by the home appliance classification model creation module.

Claims

1. An apparatus for classifying home appliances based on power consumption using deep learning, the apparatus comprising:

a home appliance classification model creation module configured to encode power consumption data collected from a customer'"'"'s house to learn a home appliance classification model and create a recurrent neural network (RNN)-based home appliance classification model; and

a home appliance classification module configured to collect and encode data on power consumption currently in use, classify home appliances using the home appliance classification model created by the home appliance classification model creation module, and output a result of the classification.

2. The apparatus of claim 1, wherein the home appliance classification model creation module comprises:

a learning data input unit configured to input the power consumption information collected from the customer'"'"'s house;

a data encoding unit configured to encode the input power consumption information;

a home appliance classification model learning unit configured to learn a home appliance classification model using the encoded data; and

a home appliance classification model creation unit configured to input new learning data and repeat the above process when a learning error is greater than or equal to a threshold and stop the learning and create a home appliance classification model when the learning error is less than or equal to the threshold.

3. The apparatus of claim 1, wherein the home appliance classification module comprises:

a power consumption information collection unit configured to collect data on power consumption currently in use;

a power consumption information input unit configured to input the collected power consumption data;

a data encoding unit configured to encode the input power consumption data;

a home appliance classification unit configured to classify home appliances based on the encoded data; and

a classification result output unit configured to output a result of the classification of the home appliances.

4. A method of classifying home appliances based on power consumption using deep learning, the method comprising:

encoding power consumption data collected from a customer'"'"'s house to learn a home appliance classification model and creating an RNN-based home appliance classification model; and

collecting and encoding data on power consumption currently in use, classifying home appliances using the RNN-based home appliance classification model, and outputting a result of the classification.

5. The method of claim 4, wherein the creating of an RNN-based home appliance classification model comprises:

inputting the power consumption information collected from the customer'"'"'s house;

encoding the input power consumption information;

learning a home appliance classification model using the encoded data; and

inputting new learning data and repeating the above process when a learning error is greater than or equal to a threshold and stopping the learning and creating a home appliance classification model when the learning error is less than or equal to the threshold.

6. The method of claim 4, wherein the classifying of home appliances and the outputting of a result of the classification comprises:

collecting data on power consumption currently in use;

inputting the collected power consumption data;

encoding the input power consumption data;

classifying home appliances using the RNN-based home appliance classification model based on the encoded data; and

outputting a result of the classification of the home appliances.

7. The method of claim 4, wherein, in the encoding of the input power consumption information included in the creating of an RNN-based home appliance classification model:

a set of the home appliances of the customer'"'"'s house is represented as Appliance;

the number of home appliances is represented as N;

when each of the home appliances in the set is represented as app₁ (1≦

i≦

N), the set of home appliances is represented as Appliance={app₁, app₂, app₃, . . . app_N}; and

when the collected power consumption is represented as P, and the maximum power consumption of each home appliance is represented as MAX(app_i), the encoded power consumption is

8. The method of claim 4, wherein the encoding of the input power consumption information included in the creating of an RNN-based home appliance classification model comprises measuring a variation while updating a base value of the power consumption with a low pass filter and encoding a value of measured total power consumption according to the variation.

9. The method of claim 4, wherein, in the encoding of the input power consumption information included in the creating of an RNN-based home appliance classification model, an output at time t is used as an input at time t+1 in order to increase a learning rate for a home appliance use pattern.

10. The method of claim 4, wherein, in the classifying of home appliances using the RNN-based home appliance classification model and the outputting of a result of the classification,

input data is composed of a previous output and data collected from a sensor installed in the customer'"'"'s house, in which the previous output is binary data that is represented as one bit stream, converted into a decimal number, and then input;

and

output data is composed of binary data that indicates an on/off state of a home appliance.

11. The method of claim 4, wherein, in the classifying of home appliances using the RNN-based home appliance classification model and the outputting of a result of the classification,

input data is power consumption collected from a sensor installed in the house or a variation of power consumption that is encoded using a low pass filter;

each output node is one of combinations of home appliances in the house; and

only one combination having the highest probability is selected without needing to set a criterion for determining on/off states when the model classifies home appliances.