Use of Machine Learning Methods  in MicroGrid

Abstract

Use of Internet of Things (IoT) concept plays the important role in modern electrotechnical systems with distributed generation. In particular, for the objects like “smart house” in the bounds of this concept some processes are considered that provide fulfillment of such control and regulation functions like following: light control; control of electrical energy consumption; monitoring of biotelemetrical parameters. Of course, this list is far from the complete one. Additionally, the range of control tasks depends on functionality and features of the system. The more is range of the tasks and number of different parameters to be measured, analyzed and processed – the more urgent become use of complicated control algorithms. Among them methods of artificial intelligent has the important place. The actuality is additionally proven by the further increasing of sales volume of the smartphones and other gadgets that combine more and more sensors and give more and more possibilities for the learning, getting and analyzing of the huge volume of different-kind data. Modern systems of energy consumption control use automated regulation of the devices and automated process of decision making. During this, demands of the end consumer get more priority because of necessity to provide admirable level of  “comfortability” while all demands of safety and life support should be answered too. However, the presence of the human  like a consumer of the services of smart house leads to the complexity of control algorithms creation due to subjectivity and  impossibility to predict human behavior. Thus, the methods that can more or less imitate human behavior in technical  system have the huge perspective. The aim of the paper is investigation of complicated intellectual methods of decision making in MicroGrid generation as well as creation of the method able to classify working regimes in MicroGrid. In machine  learning methods it is highly important to provide necessary quality of the learning based on accumulated data sets. This  quality depends not only on characteristics of dataset itself (like completeness and level of compensation) but also on   the learning technique. There is classification of the learning methods was fulfilled in the paper. Among them we considered  methods based on following concepts: symbolism; connectivity; evolution; Bayesian probability; analogies. For each method  and each ideology we pointed the idea, the concept, the approach and the most problem. Such classification allows to understand better what kind of learning is more preferable on this or that stage or in this or that situation. Then correspondence between the machine learning methods and control tasks in MicroGrid were investigated and presented. The software for the simple situation of light and heating control were elaborated and presented in the paper. Visualization of the machine learning result were executed and the areas of different classes were defined and shown.  

Keywords

References

1. L. V. Kantorovich and M. . Gavuri, “Primeneniye matematicheskikh metodov v voprosakh analiza gruzopotokov [The use of mathematical methods in the analysis of cargo flows],” Probl. povysheniya Eff. Rab. Transp., pp. 110–138, 1949.
2. D. Dantsig and F. Vol’f, “Algoritm razlozheniya dlya zadach lineynogo programmirovaniya [Decomposition Algorithm for Linear Programming Problems],” Matematika, vol. 8, no. 1, pp. 151–160, 1964.
3. O. I. Aven, S. Y. Lovetskiy, and G. Y. Moiseyenko, Optimizatsiya transportnykh potokov [Traffic optimization]. Moskva: Nauka, 1985.
4. A. V. Krylov and Y. S. Yamnenko, “Rishennya transportnoyi zadachi za dopomohoyu shtuchnoho intelektu dlya morsʹkoho transportu [Solving the transport problem using artificial intelligence for maritime transport],” Shipbuild. Mar. Infrastruct., vol. 1, no. 11, pp. 116–122, 2019, URL: http://smi.nuos.mk.ua/archive/2019/1/14.pdf.
5. V. G. Galaburda, Optimal’noye planirovaniye gruzopotokov [Optimal freight traffic planning]. Moskva: Transport, 1985.
6. A. L. Lur’ye, “Algoritm resheniya setevoy transportnoy zadachi s ogranicheniyem propusknykh sposobnostey metodom uslovno-optimal’nykh planov [An algorithm for solving a network transportation problem with bandwidth limitation by the method of conditionally optimal plans,” in Materialy Konferentsii po opytu i perspektivam primeneniya matematicheskikh metodov i EMM v planirovanii., 1962, pp. 3–13.
7. V. Y. Zhuykov et al., Pidvyshchennya efektyvnosti system z vidnovlyuvanymy dzherelamy enerhiyi [Improving the efficiency of systems with renewable energy sources]. Ky`yiv: Kafedra, 2018, ISBN: 978-617-7301-48-5.
8. G. De Jong, H. F. Gunn, and W. Walker, “National and international freight transport models: Anoverview and ideas for further development,” Transp. Rev., vol. 24, no. 1, pp. 103–124, 2004.
9. L. A. Beklaryan and N. K. Khachatryan, “Traveling wave type solutions in dynamic transport models,” Funct. Differ. Equations, vol. 13, no. 2, pp. 125–155, 2006.

License

Copyright (c) 2020 Артем Вадимович Крилов (Автор)