1. SISN
  2. All Volumes and Issues
  3. Volume 8, 2020
  4. Neural Network Method for Search of the Active Site of a Wind Power Plant

Neural Network Method for Search of the Active Site of a Wind Power Plant

SISN.
2020;
Volume 8
: pp. 55 - 64
https://doi.org/10.23939/sisn2020.08.055
Authors:
  1. Mykola Medykovskyy
  2. R. V. Melnyk
  3. Maxim Dubchak
1
Lviv Polytechnic National University, Lviv, Ukraine
2
Lviv Polytechnic National University, Lviv, Ukraine
3
Lviv Polytechnic National University

The article presents the results of the study of the possibilities of using neural networks to solve the problem of determining the active set of a wind farm (WF), taking into account the efficiency of each wind turbines (WT).

The comparative analysis of the obtained results with the known methods of determining the active composition of WF, such as: the method of dynamic programming; the method of dynamic programming with increasing the load on the experimentally set percentage; modified method of dynamic programming. The advantages and disadvantages of using each of the studied methods in terms of the possibility of achieving a given generation power at the maximum efficiency of the selected WT are determined.

It is established that when using recurrent neural networks to solve the problem of determining the active composition of WF, the minimum direct linear variation of the difference between the power to be generated and the actual power of the determined active set of WF is 2.7%. Under the same conditions, the use of other known methods, in particular, the modified method of dynamic programming ensures the achievement of this parameter at the level of 0.05%. This significantly increases the time to solve the problem. By computer simulation, it was found that under equal conditions, the time to solve the problem using neural networks — 0.04 s, and using a modified method of dynamic programming 3.4 s. The obtained results provide an opportunity to implement effective decision support systems in energy flow management.

Wind Power Plant
artificial neural networks
active set of wind power plant
optimization
  1. Medykovskyy M., Shunevich O. (2010). Multicriteria method for evaluating the efficiency of a wind power plant. Kyiv: Bulletin of the Engineering Academy of Ukraine 4. 240-245.
  2. Tyuptya V.I., Shevchenko V.I., Stryuk V.K. (2003). Dynamic and nonlinear programming. Kyiv: Electronic Library of the Faculty of Cybernetics
  3. Medikovsky M., Shunevich O. (2010). Using integer programming to determine the composition of a wind farm. Kyiv: Modeling and Information Technologies 57. 230-233.
  4. Medykovskyy M., Shunevich O. (2009). Method for determining the structure of a wind power plant taking into account the load dynamics. Kyiv: Modeling and information technologies. 175-181.
  5. Shunevich O. (2013). Information technology of formation of dynamic composition of wind power plant. (Abstract of the dissertation for the degree of candidate of technical sciences). Lviv Polytechnic National University. Lviv [in Ukrainian].
  6. Kravchyshyn V., Medykovskyy M., Melnyk R. (2016). Modification of Dynamic Programming Method in Determining Active Composition of Wind Power Stations. Computational problems of electrical engineering Vol. 6, 83-90.
  7. Teslyuk T.V., Tsmots I.G., Teslyuk V.M., Medykovskyy M.O. (2017). Optimization of the structure of a wind power plant using the method of branches and boundaries. Eastern European Journal of Advanced Technologies 2/8 (86).
    https://doi.org/10.15587/1729-4061.2017.96769
  8. Yoshua Bengio, Andrea Lodi, Antoine Prouvost (2018). Machine Learning for Combinatorial Optimization: Methodological Tour d'Horizon. Retrieved from https://arxiv.org/abs/1811.06128 [in English].
  9. Kate A. Smith (1999). Neural Networks for Combinatorial Optimization: A Review of More Than a Decade of Research. Informs Journal on Computing Vol. 11(1). 15-34.
    https://doi.org/10.1287/ijoc.11.1.15
  10. Shigeo Abe, Junzo Kawakami, Kotaroo Hirasawa (1992). Solving Inequality Constrained Combinatorial Optimization Problems by the Hopfield. Neural Networks Vol. 5(4). 663-670.
    https://doi.org/10.1016/S0893-6080(05)80043-7
  11. Davide Martini (2012). Application of Neural Network for the Knapsack Problem. (Magister theses). Università degli studi di padova. Padua.
  12. J. Deane and Anurag Agarwal (2012). Neural metaheuristics for the multidimensional knapsack problem. Technical report.
  13. Mattias Ohlsson, Carsten Peterson, Bo Söderberg (1993). Neural Networks for Optimization Problems with Inequality Constraints: The Knapsack Problem. Mainz Institute for Theoretical Physics Vol. 5. 331-339.
    https://doi.org/10.1162/neco.1993.5.2.331
  14. Ben Krause, Liang Lu, Iain Murray, Steve Renals (2015). On the Efficiency of Recurrent Neural Network Optimization Algorithms. NIPS Optimization for Machine Learning Workshop.
  15. Jianli Feng, Shengnan Lu, (2019). Performance Analysis of Various Activation Functions in Artificial Neural Networks. Journal of Physics Conference Series.
    https://doi.org/10.1088/1742-6596/1237/2/022030
  16. A. Trask, F. Hill, S. E. Reed, J. Rae, C. Dyer, P. Blunsom (2018). Neural arithmetic logic units. NIPS'18: Proceedings of the 32nd International Conference on Neural Information Processing Systems. 8035-8044.
  17. Enercon E-53 - 800,00 kW - Wind turbine. Retrieved from https://en.wind-turbine- models.com/turbines/530-enercon-e-53 [in English].
  18. Enercon E-44 - 900,00 kW - Wind turbine. Retrieved from: https://en.wind-turbine- models.com/turbines/531-enercon-e-44 [in English].