• 沒有找到結果。

Discussion on CS-UP Mining

Usage Alertion

6.1 Data Sets

6.2.3 Discussion on CS-UP Mining

We know the CS-UP is improved method for S-UP. Therefore, it also has minimum support ρ, time extraordinariness threshold α, and duration extraordinariness threshold β. At first, the setting of ρ is the same as S-UP and we also extract the CS-UP. At first, the setting of ρ is the same as S-UP. The Table 6.5 shows the normal CS-UP for each appliance. Then, we discuss with setting α and β. As shown in Figure 6.6a and 6.6b, we can see all of the appliances have in the same range of α. The range of α is from 0.4 hours to 1 hours. For the β, only the β range of microwave is from 2 minutes to 7 minutes and the others are from 2 minutes to 32 minutes. Because we use the clustering method to make each month jth time start time and duration close to the normal behavior, the range of time extraordinariness threshold and duration extraordinariness isn’t dispersion. The Table 6.6 shows the extraordinary CS-UP for each appliance.

# of extraordinary usage patterns # of extraordinary usage patterns

time extraordinariness threshold α (hours)

microwave

# of extraordinary usage patterns # of extraordinary usage patterns

duration extraordinariness threshold β (hours)

microwave

Figure 6.6: Relation between extraordinary threshold and number of extraordinary usage patterns for CS-UP

Table 6.6: Extraordinary CS-UP for each appliance at α = 0.4 hours and β = 2 minutes

appliance Name centroid of cluster

jth time mean of time mean of duration(hours)

microwave

1 02:50:00 0.03

2 04:20:12 0.05

3 06:32:22 0.07

4 06:35:47 0.05

5 06:48:21 0.03

light

1 02:50:00 0.03

2 07:30:12 12.03

3 20:32:22 0.2

4 21:27:11 1.1

wash-dryer 1 00:07:20 0.6

2 23:30:13 0.48

dish-washer 1 09:00:04 2.1

oven

1 06:59:23 0.68

2 22:30:10 0.3

3 23:05:31 0.23

air-conditioner

1 00:10:14 11.2

2 22:07:13 2.01

3 23:10:00 1.67

and it has eight different colors of curve. In other words, it has eight different representative daily usage behavior. Then, we discuss with setting parameter γ. The Figure 6.8 shows the relation between γ and the number of extraordinary usage patterns and the Figure 6.9 shows the one of extraordinary DB-UPs for each appliance.

0 1

00:00 03:00 06:00 09:00 12:00 15:00 18:00 21:00 00:00

ON/OFF

00:00 03:00 06:00 09:00 12:00 15:00 18:00 21:00 00:00

ON/OFF

00:00 03:00 06:00 09:00 12:00 15:00 18:00 21:00 00:00

ON/OFF

00:00 03:00 06:00 09:00 12:00 15:00 18:00 21:00 00:00

ON/OFF

00:00 03:00 06:00 09:00 12:00 15:00 18:00 21:00 00:00

ON/OFF

00:00 03:00 06:00 09:00 12:00 15:00 18:00 21:00 00:00

ON/OFF

0

2500 2200 1700 1200 700 200

0

# of extraordinary usage patterns # of extraordinary usage patterns

extraordinary threshold γ (hours)

microwave

Figure 6.8: Relation between extraordinary threshold and number of extraordinary usage patterns for DB-UP

0 1

00:0002:00 04:0006:0008:00 10:0012:0014:00 16:0018:0020:00 22:0000:00

ON/OFF

00:0002:0004:00 06:0008:0010:00 12:0014:0016:00 18:0020:0022:00 00:00

ON/OFF

00:0002:00 04:0006:0008:00 10:0012:0014:00 16:0018:0020:00 22:0000:00

ON/OFF

00:0002:0004:00 06:0008:0010:00 12:0014:0016:00 18:0020:0022:00 00:00

ON/OFF

00:0002:00 04:0006:0008:00 10:0012:0014:00 16:0018:0020:00 22:0000:00

ON/OFF

00:0002:0004:00 06:0008:0010:00 12:0014:0016:00 18:0020:0022:00 00:00

ON/OFF

time

Air-Conditioned

(f) Air-conditioner

Figure 6.9: Extraordinary DB-UP for each appliance at γ = 200

Chapter 7 Conclusion

In this paper, we propose a system HAUBA to provide appliance usage information. For our system, we propose four methods to represent appliance usage behavior and identify what the extraordinary usage behavior is for each usage pattern. For the four types of usage pattern, the first is time slot usage pattern (TS-UP). We can know when the appliance is be highly used from time slot usage pattern, but we can’t know the exact time and duration. Therefore, the statistical usage pattern (S-UP) discuss in detail. It extract each time to turn on the appliance and evaluate the mean of the start time and mean of the duration for each time.

The S-UP provide the information about average of start time and average of duration for the jth time to turn on the appliance, but we discover that there are different start time and duration for the jth time to turn on the appliance every day. In other words, there are the different behaviors of the jth time to turn on the appliance every day. Furthermore, the noise data also appears in the jth time to turn on the appliance. Based on two problems for S-UP, we propose the clustered-based statistical usage pattern (CS-UP)to solve them. The CS-UP find the similar start time and duration (similar behavior) to group them and find the maximum group to represent the behavior of the jth time to turn on the appliance. Therefore, we can get accurate mean of time and mean of duration to represent the j time to turn

Bibliography

[1] O. Aritoni and V. Negru. A methodology for household appliances behaviour recognition in ami systems. In Proceedings of the 7th International Conference on Autonomic and Autonomous Systems (ICAS’11), pages 175–178, 2011.

[2] J. Aßfalg, H. Kriegel, P. Kr¨oger, P. Kunath, A. Pryakhin, and M. Renz. Similarity search on time series based on threshold queries. EDBT, 3896:276–294, 2006.

[3] D. Berndt and J. Clifford. Using dynamic time warping to find patterns in time series. In Proceedings of KDD-94: AAAI Workshop on Knowledge Discovery in Databases, pages 359–370, 1994.

[4] L. Chen and R. Ng. On the marriage of lp-norms and edit distance. In Proceedings of the 30th International Conference on Very Large Data Bases (VLDB’04), volume 30, pages 792–803, 2004.

[5] L. Chen, M. ¨Ozsu, and V. Oria. Robust and fast similarity search for moving object trajectories. In Proceedings of the 31st ACM SIGMOD international conference on Man-agement of data (SIGMOD’05), pages 491–502, 2005.

[6] Y. Chen, M. Nascimento, B. Ooi, and A. Tung. Spade: On shape-based pattern detection in streaming time series. In Proceedings of the 23rd International Conference on Data Engineering( ICDE’07), pages 786–795, 2007.

[7] H. Ding, G. Trajcevski, P. Scheuermann, X. Wang, and E. Keogh. Querying and mining of time series data: experimental comparison of representations and distance measures.

Proceedings of the 34th International Conference on Very Large Data Bases (VLDB’08), 1:1542–1552, 2008.

[8] L. Farinaccio and R. Zmeureanu. Using a pattern recognition approach to disaggregate the total electricity consumption in a house into the major end-uses. Energy and Buildings, 30(3):245–259, 1999.

[9] E. Frentzos, K. Gratsias, and Y. Theodoridis. Index-based most similar trajectory search.

In Proceedings of the 23rd International Conference on Data Engineering (ICDE’07), pages 816–825, 2007.

[10] H. Goncalves, A. Ocneanu, and M. Berg´es. Unsupervised disaggregation of appliances using aggregated consumption data. In Proceedings of the 17th ACM SIGKDD workshop on Data Mining Applications for Sustainability (SustKDD’11), 2011.

[11] G. Hart. Nonintrusive appliance load monitoring. Proceedings of the IEEE, 80(12):1870–

1891, 1992.

[12] J. Huang, H. Sun, J. Han, H. Deng, Y. Sun, and Y. Liu. Shrink: a structural clustering algorithm for detecting hierarchical communities in networks. In Proceedings of the 19th ACM International Conference on Information and Knowledge Management (CIKM’10, pages 219–228, 2010.

[13] M. Ito, R. Uda, S. Ichimura, K. Tago, T. Hoshi, and Y. Matsushita. A method of appliance detection based on features of power waveform. In International Symposium on Applications and the Internet (SAINT’04), pages 291–294, 2004.

[14] T. Kato, H. Cho, D. Lee, T. Toyomura, and T. Yamazaki. Appliance recognition from electric current signals for information-energy integrated network in home environments.

pages 150–157, 2009.

[15] H. Kim, M. Marwah, M. Arlitt, G. Lyon, and J. Han. Unsupervised disaggregation of low frequency power measurements. In Proceedings of the 11th SIAM International

[17] G. Lin, S. Lee, J. Hsu, and W. Jih. Applying power meters for appliance recognition on the electric panel. In Proceedings of the 5th IEEE Conference on Industrial Electronics and Applications (ICIEA’10), pages 2254–2259, 2010.

[18] H. Matthews, L. Soibelman, M. Berges, and E. Goldman. Automatically disaggregat-ing the total electrical load in residential builddisaggregat-ings: a profile of the required solution.

Intelligent Computing in Engineering, pages 381–389, 2008.

[19] A. Prudenzi. A neuron nets based procedure for identifying domestic appliances pattern-of-use from energy recordings at meter panel. volume 2, pages 941–946, 2002.

[20] K. Suzuki, S. Inagaki, T. Suzuki, H. Nakamura, and K. Ito. Nonintrusive appliance load monitoring based on integer programming. In SICE Annual Conference, pages 2742–2747, 2008.

[21] M. Vlachos, G. Kollios, and D. Gunopulos. Discovering similar multidimensional trajecto-ries. In Proceedings of the 18th International Conference on Data Engineering(ICDE’02), pages 673–684, 2002.

相關文件