![](bg8.png)
Table 5: Results obtained for load classification using LSTM and CNN.
Network Heater Drain Pump Electrovalves
LSTM
F1 class ON 99.82% 97.44% 37.61%
F1 class OFF 99.68% 99.54% 87.81%
CNN
F1 class ON 99.72% 98.57% 98.39%
F1 class OFF 99.95% 99.75% 99.87%
obtained for the drain pump. This load is quite easy
to detect and hence both the deep learning approaches
provide good results. The same cannot be said for
electrovalves. Regarding this load, it is clear that
CNNs outperform LSTMs.
6 CONCLUSIONS
In this work two different problems have been faced:
the drum speed estimation of a washing machine and
the activation status classification of different loads of
the same appliance. The first has been solved training
an LSTM network that estimates the speed at each
time instant. Results on the test set prove that good
performances can be achieved using this network, es-
pecially if the state dimension of the network is set
solving an optimization problem.
As for the second problem, two different ap-
proaches have been tested. The first consisted in
training an LSTM network (with an optimal number
of hidden units) whereas the second makes use of
CNNs. Good results have been achieved using both
the networks for two out of three loads (heater and
drain pump). Conversely, it is clear that using only
a weighted classification layer in the electrovalves-
status classification, is not enough to cope with class
unbalance, thus using CNNs leads to much better re-
sults. Hence, even though LSTMs are easier to train
and test (since the only preprocessing operation re-
quired is the normalization), CNNs will be preferred
since perform better in classifying all the loads.
REFERENCES
Alasalmi, T., Suutala, J., and R
¨
oning, J. (2012). Real-
time non-intrusive appliance load monitor. In Inter-
national conference on Smart grids and Green IT Sys-
tems, pages 203–208.
Batista, G. E., Prati, R. C., and Monard, M. C. (2004). A
study of the behavior of several methods for balancing
machine learning training data. ACM SIGKDD explo-
rations newsletter, 6(1):20–29.
Buda, M., Maki, A., and Mazurowski, M. A. (2018). A
systematic study of the class imbalance problem in
convolutional neural networks. Neural Networks,
106:249–259.
Djordjevic, S. and Simic, M. (2018). Nonintrusive identifi-
cation of residential appliances using harmonic anal-
ysis. Turkish Journal of Electrical Engineering and
Computer Sciences, 26.
Glorot, X., Bordes, A., and Bengio, Y. (2011). Deep sparse
rectifier neural networks. In Proceedings of the four-
teenth international conference on artificial intelli-
gence and statistics, pages 315–323.
Goodfellow, I., Bengio, Y., and Courville, A. (2016). Deep
learning. MIT press.
Grangier, D., Bottou, L., and Collobert, R. (2009). Deep
convolutional networks for scene parsing. In ICML
2009 Deep Learning Workshop, volume 3, page 109.
Citeseer.
Hannun, A. Y., Rajpurkar, P., Haghpanahi, M., Tison, G. H.,
Bourn, C., Turakhia, M. P., and Ng, A. Y. (2019).
Cardiologist-level arrhythmia detection and classifi-
cation in ambulatory electrocardiograms using a deep
neural network. Nature medicine, 25(1):65.
Hochreiter, S. and Schmidhuber, J. (1997). Long short-term
memory. Neural computation, 9(8):1735–1780.
Huang, X., Yin, B., Zhang, R., and Wei, Z. (2019). Study
of steady-state feature extraction algorithm based on
emd. In IOP Conference Series: Materials Science
and Engineering, volume 490, page 062036. IOP Pub-
lishing.
Jeni, L. A., Cohn, J. F., and De La Torre, F. (2013). Fac-
ing imbalanced data–recommendations for the use of
performance metrics. In 2013 Humaine association
conference on affective computing and intelligent in-
teraction, pages 245–251. IEEE.
Kim, J., Le, T.-T.-H., and Kim, H. (2017). Nonintrusive
load monitoring based on advanced deep learning and
novel signature. Computational intelligence and neu-
roscience, 2017.
Krizhevsky, A., Sutskever, I., and Hinton, G. E. (2012). Im-
agenet classification with deep convolutional neural
networks. In Advances in neural information process-
ing systems, pages 1097–1105.
Maitre, J., Glon, G., Gaboury, S., Bouchard, B., and
Bouzouane, A. (2015). Efficient appliances recog-
nition in smart homes based on active and reactive
power, fast fourier transform and decision trees. In
Workshops at the Twenty-Ninth AAAI Conference on
Artificial Intelligence.
Mocanu, E., Nguyen, P. H., Gibescu, M., and Kling, W. L.
(2016). Deep learning for estimating building energy
consumption. Sustainable Energy, Grids and Net-
works, 6:91–99.
Susto, G. A., Zambonin, G., Altinier, F., Pesavento, E., and
Beghi, A. (2018). A soft sensing approach for clothes
load estimation in consumer washing machines. In
2018 IEEE Conference on Control Technology and
Applications (CCTA), pages 1252–1257. IEEE.
ICPRAM 2020 - 9th International Conference on Pattern Recognition Applications and Methods
432