Machine Learning Methods for Forecasting Electricity Consumption of Prosumers
DOI:
https://doi.org/10.71701/fzkw9s31Keywords:
Prosumer, electricity consumption forecasting, k-Nearest Neighbors, Artificial Neural Networks, machine learningAbstract
The trend towards the so-called digitalization of energy systems brings with it the implementation of smart meters and the development of the “prosumer” market (energy producers and consumers which generate electricity locally), which demands emerging technologies, such as machine learning, to improve energy management. This research aims to build models based on machine learning methods for forecasting hourly electricity consumption of a residential prosumer located in Germany. In this study, a methodology is developed to build prediction models based on k-Nearest Neighbors and Artificial Neural Networks methods, which are applied on historical electricity consumption data obtained along fifteen months and in combination with local temperature data. Despite the complexity of predicting consumption due to the irregular usage of electricity in a household, the k-Nearest Neighbors and Artificial Neural Networks models showed acceptable accuracy results with a medium absolute percentage error around 30% in three different scenarios with forecast periods of 48, 24 and 1 hours. The predictive models can be implemented through emerging prosumer-oriented business models, whose value proposition aims at reducing electricity costs through improvement of renewable energy selfconsumption.
Downloads
References
DWD Climate Data Center (2018). Historical hourly station observations of 2m air temperature and humidity for Germany. Version v006. Recuperado de ftp://opendata.dwd.de/climate_environment/CDC/observations_germany/climate/hourly/.
Fan, C., Xiao, F., Wang, S. (2014). Development of prediction models for next-day building energy consumption and peak power demand using data mining techniques. Applied Energy, 127, 1–10.
Gajowniczek, K., Ząbkowski, T. (2017). Electricity forecasting on the individual household level enhanced based on activity patterns. PloS one, 12 (4). Recuperado de https://doi.org/10.1371/journal.pone.0174098.
He, W. (2008). Forecasting electricity load with optimized local learning models. International Journal of Electrical Power & Energy Systems, 30 (10), 603–608.
International Energy Agency (2017). Digitalization and Energy. International Energy Agency (IEA). Recuperado de https://www.iea.org/publications/freepublications/publication/DigitalizationandEnergy3.pdf.
Kotu, V., Deshpande, B. (2015). Predictive analytics and data mining. Concepts and practice with RapidMiner. Amsterdam: Elsevier/Morgan Kaufmann.
Kuster, C., Rezgui, Y., Mourshed, M. (2017). Electrical load forecasting models. A critical systematic review. In Sustainable Cities and Society, 35, 257–270.
Makridakis, S., Spiliotis, E., Assimakopoulos, Vassilios (2018). Statistical and Machine Learning forecasting methods. Concerns and ways forward. PloS one, 13 (3). Recuperado de https://doi.org/10.1371/journal.pone.0194889.
Moon, J., Park, J., Hwang, E., Jun, S. (2018). Forecasting power consumption for higher educational institutions based on machine learning. The Journal of Supercomputing, 74 (8), 3778–3800.
Open Power System Data (2017). Data Package Household Data. Versión 2017-11-10. Recuperado de https://doi.org/10.25832/household_data/2017-11-10.
RapidMiner Inc. (2018). RapidMiner Studio. Versión 9.0.3. Recuperado de https://rapidminer.com/getstarted.
Shalev-Shwartz, S., Ben-David, S. (Eds.) (2014). Understanding machine learning. From theory to algorithms. New York: Cambridge University Press.
Suthaharan, S. (2016). Machine Learning Models and Algorithms for Big Data Classification. Boston, MA:
Springer US.
Valgaev, O., Kupzog, F., Schmeck, Harmut (2016). Low-voltage power demand forecasting using K-nearest neighbors approach. 2016 IEEE Innovative Smart Grid Technologies - Asia (ISGT-Asia). Melbourne: IEEE, 1019–1024.
Zhang, R., Xu, Y., Dong, Z., Kong, W., Wong, K. (2016). A composite k-nearest neighbor model for day-ahead load forecasting with limited temperature forecasts. 2016 IEEE Power and Energy Society General Meeting (PESGM). Boston: IEEE, 1–5.
Zufferey, T., Ulbig, A., Koch, S., Hug, G. (Eds.) (2017). Forecasting of Smart Meter Time Series Based on Neural Networks. Data Analytics for Renewable Energy Integration. Cham: Springer International Publishing.
Downloads
Published
Issue
Section
License
This work is licensed under a Creative Commons Attribution-NonCommercial 4.0 International License.
