Comparison of neural network architectures to predicting electricity consumption by enterpris

DOI: 10.47026/1810-1909-2023-4-57-65

УДК 621.31

ББК 31.2

Denis V. BORTNIK, Aleksandr I. ORLOV

Key words

neural network, recurrent neural networks, WaveNet, prediction, one-dimensional convolutional networks

Abstract

Forecasting of electricity consumption is a key tool for enterprises, energy supply and power grid organizations. Accurate forecasting enables to plan the distribution of limited resources of the power grid facilities, as well as to manage the demand for electricity. In the context of modern demand management technologies, improving the accuracy of forecasting of electricity consumption becomes especially important.

The purpose of the study is to improve the accuracy of predicting power consumption by the power supply object using neural networks.

Materials and methods. The work used a data set containing a profile of the enterprise’s capacity for a three-month period, as well as additional data, such as time of day, day of the week, weekends and holidays, month. The data set is divided into training and control parts. Preliminary data processing, neural network architecture design, training and testing were carried out. The criteria for the quality of training are the mean absolute error, the mean square error and the coefficient of determination.

Research results. In the work, a comparative analysis of three neural network architectures was performed: a one-dimensional convolutional network, a recurrent neural network with long-term and short-term memory, and WaveNet, on the basis of which their indicators of the quality of power consumption forecasting were evaluated. It was shown that all considered architectures of neural networks are suitable for the use in the issue of predicting power consumption. Long-term and short-term memory networks have shown good results in power prediction due to their ability to handle long-term time dependencies. The WaveNet architecture outperformed both long-term and short-term memory model-based recurrent neural networks and one-dimensional convolutional networks by selected criteria.

Conclusions. The study led to the conclusion that the use of neural networks, especially architectures with long-term and short-term memory and WaveNet, is an effective approach for predicting power consumption. The quality of forecasting significantly depends on the choice of hyperparameters and preliminary processing of the initial data. Prospect for further research in this area is studying the influence of various factors on the accuracy of forecasting and optimization of the learning process of neural networks.

References

1. Alekseeva I.Yu. Kratkosrochnoe prognozirovanie elektropotrebleniia v elektroenergeticheskikh sistemakh s ispol’zovaniem iskusstvennykh neironnykh setei: avtoref. dis. … kand. tekhn. nauk [Short-term forecasting of energy consumption in the electrical systems using artificial neural networks. Abstract of Cand. Diss.]. Ivanovo, 2014. P. 9–10.
2. Arzamastsev A.A., Kriuchin O.V., Azarova P.A., Zenkova N.A. Universal’nyi programmnyi kompleks dlia komp’iuternogo modelirovaniia na osnove iskusstvennoi neironnoi seti s samoorganizatsiei struktury g. Tambov [Universal software complex for computer modeling based on artificial neural network with self-organizational structure in Tambov]. Vestnik Tambovskogo universiteta. Ser.Estestvennye nauki i tekhnicheskie nauki, 2006, vol. 11, iss. 5, pp. 372–375.
3. Vyalkova S.A., Kornyukova O.A., Nadtoka I.I. Razrabotka modelei kratkosrochnogo prognozirovaniya sutochnykh grafikov aktivnoi moshchnosti g. Moscow. [Development of models for short-term forecasting of daily schedules of active power in Moscow.]. In: Problemy mashinovedeniya: materialy V Mezhdunar. nauch.-tekhn. konf. [Proc. of the V Int. Sci. and Tech. Conf. «Problems of machine science»]. Omsk, 2021, pp. 166–
4. Vyalkova S.A., Nadtoka I.I. Prognozirovanie sutochnykh grafikov aktivnoi moshchnosti megapolisa s uchetom prognoznykh dannykh estestvennoi osveshchennosti [Forecasting daily schedules of the active power of the metropolis, taking into account the predictive data of natural illumination]. Izvestiya vysshikh uchebnykh zavedenii. Elektromekhanika, 2020, vol. 63, no. 5, pp. 67–71.
5. Akhmet’ianov R.R., Delegodina L.A., Kopylova N.P. et al. Mul’tiplikativnaia model’ sezonnogo energopotrebleniia predpriiatii [Multiplicative model of seasonal energy consumption at enterprises]. Avtometriya, 2008, vol. 44, no. 3, pp. 106–118.
6. Nadtoka I.I., Al’ Zikheri B.M. Baiesovskaia neirosetevaia model’ dlia kratkosrochnogo prognozirovaniia elektropotrebleniia regiona g. Novocherkassk [Bayes neural network model for short-term forecasting energy consumption in the region in Novocherkassk]. In: Studencheskaia nauchnaia vesna: Materialy regional’noi nauchno-prakticheskoi konferentsii studentov, aspirantov i molodykh uchenykh vuzov Rostovskoi oblasti [Proc. of Sci. Conf. «Student Scientific Spring»]. Novocherkassk, 2012, pp. 136–138.
7. Aurélien G. Hands-on machine learning with Scikit-Learn, Keras, and TensorFlow Bejing [etc.]: O`Reilly, cop. 2020, 856 p. (Russ. ed.: Prikladnoe mashinnoe obuchenie s pomoshch’yu Scikit-Learn, Keras i TensorFlow: kontseptsii, instrumenty i tekhniki dlya sozdaniya intellektual’nykh system. 2-e izd. Petersburg, Dialektika Publ., 2020, 1040 p.).
8. Solov’eva I.A., Dziuba A.P. Prognozirovanie elektropotrebleniia s uchetom faktorov tekhnologicheskoi i rynochnoi sredy [Energy consumption forecasting with the factors of technological and market environment]. Nauchnyi dialog, 2013, no. 7 (19), pp. 97–111.
9. Aileen N. Practical Time Series Analysis: Prediction with Statistics and Machine Learning, O’Reilly Media; 1^st (November 5, 2019), 504 p.
10. Chollet F. Deep Learning with Python. 2^nd Manning Publications, 2021, 384 p.
11. Hyndman R.J., Athanasopoulos G. Forecasting: principles and practice. Publisher: OTexts, 2013. 291 p.
12. Lazzeri F. Machine Learning for Time Series Forecasting with Python. 1^st Wiley, 2020, 224 p.
13. Ruey S.T. Multivariate Time Series Analysis: With R and Financial Applications. 1^st Wiley, 2013, 520 p.
14. Shayeghi H. STLF Based on Optimized Neural Network Using PSO // International Journal of Electrical and Computer Engineering. 2009. 4:10. P. 1190–1199.
15. Zelinka I., Chen G., Rössler O.E., Snášel V., Abraham A. Nostradamus 2013: Prediction, Modeling and Analysis of Complex Systems. Springer International Publishing, Switzerland, 2013. 550 p.

Information about the authors

Denis V. Bortnik – Post-Graduate Student, Department of Power Supply and Technical Diagnostics, Mari State University, Russia, Yoshkar-Ola (bortnik_denis16@mail.ru; ORCID: https://orcid.org/0009-0002-7010-8271).

Aleksandr I. Orlov – Candidate of Technical Sciences, Head of the Department of Electromechanics, Mari State University, Russia, Yoshkar-Ola (karlorlov@gmail.com; ORCID: https://orcid.org/0000-0003-1152-6668).

For citations

Bortnik D.V., Orlov A.I. COMPARISON OF NEURAL NETWORK ARCHITECTURES TO PREDICTING ELECTRICITY CONSUMPTION BY ENTERPRISE. Vestnik Chuvashskogo universiteta, 2023, no. 4, pp. 57–65. DOI: 10.47026/1810-1909-2023-4-57-65 (in Russian).

Download the full article