Gradient Boosting Models for Photovoltaic Power Estimation Under Partial Shading Conditions
The energy yield estimation of a photovoltaic (PV) system operating under partially shaded conditions is a challenging task and a very active area of research. In this paper, we attack this problem with the aid of machine learning techniques. Using data simulated by the equivalent circuit of a PV string operating under partial shading, we train and evaluate three different gradient boosted regression tree models to predict the global maximum power point (MPP). Our results show that all three approaches improve upon the state-of-the-art closed-form estimates, in terms of both average and worst-case performance. Moreover, we show that even a small number of training examples is sufficient to achieve improved global MPP estimation. The methods proposed are fast to train and deploy and allow for further improvements in performance should more computational resources be available.
KeywordsGradient boosting Solar energy Photovoltaic (PV) system Maximum power point (MPP) Partial shading Machine learning
This project was partially supported by the EPSRC Centre for Doctoral Training [EP/I028099/1] & the EPSRC LAMBDA [EP/N035127/1] & Anyscale Apps [EP/L000725/1] project grants. N. Nikolaou acknowledges the support of the EPSRC Doctoral Prize Fellowship. E. Batzelis carried out this research at NTUA, Athens, Greece under the support of the ‘IKY Fellowships of Excellence for Postgraduate Studies in Greece-Siemens Program’.
- 1.Batzelis, E., Kampitsis, G.E., Papathanassiou, S.A.: A MPPT algorithm for partial shading conditions employing curve fitting. In: EU PVSEC, pp. 1502–1507 (2016)Google Scholar
- 5.Busa-Fekete, R., Kégl, B., Éltető, T., Szarvas, G.: Ranking by calibrated AdaBoost. In: Proceedings of the Learning to Rank Challenge, pp. 37–48 (2011)Google Scholar
- 6.Caruana, R., Niculescu-Mizil, A.: An empirical comparison of supervised learning algorithms. In: ICML, pp. 161–168 (2006)Google Scholar
- 7.Chen, T., Guestrin, C.: Xgboost: a scalable tree boosting system. In: SIGKDD, pp. 785–794 (2016)Google Scholar
- 9.Dolan, J.A., Lee, E., Yeh, C., Ben-Menahem, S., Ishihara, A.K.: Neural network estimation of photovoltaic IV curves under partially shaded conditions. In: IJCNN, pp. 1358–1365 (2011)Google Scholar
- 12.Mason, L., Baxter, J., Bartlett, P., Frean, M.: Boosting algorithms as gradient descent. In: NIPS, pp. 512–518. MIT Press (2000)Google Scholar
- 13.Moballegh, S., Jiang, J.: Modeling, prediction, and experimental validations of power peaks of PV arrays under partial shading conditions. Sustain. Energy 5(1), 293–300 (2014)Google Scholar
- 14.Nguyen, D.D., Lehman, B., Kamarthi, S.: Performance evaluation of solar photovoltaic arrays including shadow effects using neural network. IEEE Energy Convers. Congr. Expo. 6(2), 3357–3362 (2009)Google Scholar
- 16.Psarros, G., Batzelis, E., Papathanassiou, S.: Analysis of local MPPs on the P-V curve of a partially shaded PV string. In: EU PVSEC, pp. 3383–3389 (2014)Google Scholar
- 20.Wyner, A.J., Olson, M., Bleich, J., Mease, D.: Explaining the success of adaboost and random forests as interpolating classifiers (2017). arXiv:1504.07676v2