Forecasting of the performance of the shipboard electric power system on the basis of the artificial neural network
To date, the main limiting factor in development of forecasting systems based on mathematical methods of data processing, which in most cases is reduced to solving linear deterministic multiparameter problems, is the performance of a computer. Therefore, considerable attention is paid to development...
Main Authors: | , , |
---|---|
Format: | Article |
Language: | English |
Published: |
PC Technology Center
2017-07-01
|
Series: | Tehnologìčnij Audit ta Rezervi Virobnictva |
Subjects: | |
Online Access: | http://journals.uran.ua/tarp/article/view/108528 |
id |
doaj-f46140c2784f4d11ad872d37dcb80de8 |
---|---|
record_format |
Article |
spelling |
doaj-f46140c2784f4d11ad872d37dcb80de82020-11-25T01:36:18ZengPC Technology CenterTehnologìčnij Audit ta Rezervi Virobnictva2226-37802312-83722017-07-0141(36)434910.15587/2312-8372.2017.108528108528Forecasting of the performance of the shipboard electric power system on the basis of the artificial neural networkIrina Gvozdeva0Valery Lukovtsev1Sergii Tierielnyk2National University «Odessa Maritime Academy», 8, Didrikhsona str., Odessa, Ukraine, 65029National University «Odessa Maritime Academy», 8, Didrikhsona str., Odessa, Ukraine, 65029National University «Odessa Maritime Academy», 8, Didrikhsona str., Odessa, Ukraine, 65029To date, the main limiting factor in development of forecasting systems based on mathematical methods of data processing, which in most cases is reduced to solving linear deterministic multiparameter problems, is the performance of a computer. Therefore, considerable attention is paid to development and research of neural network methods for solving such problems, which is explained by the inherent massively parallel processing of information that allows building high-performance computing systems. In connection with this, the aim of this work is development of a system for predicting the SEPS performance on the basis of an artificial neural network implemented by the architecture of a multilayer perceptron. The problem of parameter normalization is solved, caused by the fact that the SEPS mode is characterized by parameters of different physical nature that have different dimensions. The task of training an artificial neural network is also solved. As a learning method, the back propagation algorithm is chosen. For the formation of a rational training sample used in the learning of an artificial neural network, mathematical methods of temporary extrapolation are used. The analysis of the obtained results shows that the value of the mean absolute error on the test set is 3.8 %. This allows to judge the possibility of using an artificial neural network to solve the problems of predicting the SEPS state.http://journals.uran.ua/tarp/article/view/108528forecasting of the state of the shipboard electric power systemcoefficient of the generalized parameterartificial neural network |
collection |
DOAJ |
language |
English |
format |
Article |
sources |
DOAJ |
author |
Irina Gvozdeva Valery Lukovtsev Sergii Tierielnyk |
spellingShingle |
Irina Gvozdeva Valery Lukovtsev Sergii Tierielnyk Forecasting of the performance of the shipboard electric power system on the basis of the artificial neural network Tehnologìčnij Audit ta Rezervi Virobnictva forecasting of the state of the shipboard electric power system coefficient of the generalized parameter artificial neural network |
author_facet |
Irina Gvozdeva Valery Lukovtsev Sergii Tierielnyk |
author_sort |
Irina Gvozdeva |
title |
Forecasting of the performance of the shipboard electric power system on the basis of the artificial neural network |
title_short |
Forecasting of the performance of the shipboard electric power system on the basis of the artificial neural network |
title_full |
Forecasting of the performance of the shipboard electric power system on the basis of the artificial neural network |
title_fullStr |
Forecasting of the performance of the shipboard electric power system on the basis of the artificial neural network |
title_full_unstemmed |
Forecasting of the performance of the shipboard electric power system on the basis of the artificial neural network |
title_sort |
forecasting of the performance of the shipboard electric power system on the basis of the artificial neural network |
publisher |
PC Technology Center |
series |
Tehnologìčnij Audit ta Rezervi Virobnictva |
issn |
2226-3780 2312-8372 |
publishDate |
2017-07-01 |
description |
To date, the main limiting factor in development of forecasting systems based on mathematical methods of data processing, which in most cases is reduced to solving linear deterministic multiparameter problems, is the performance of a computer. Therefore, considerable attention is paid to development and research of neural network methods for solving such problems, which is explained by the inherent massively parallel processing of information that allows building high-performance computing systems.
In connection with this, the aim of this work is development of a system for predicting the SEPS performance on the basis of an artificial neural network implemented by the architecture of a multilayer perceptron. The problem of parameter normalization is solved, caused by the fact that the SEPS mode is characterized by parameters of different physical nature that have different dimensions. The task of training an artificial neural network is also solved. As a learning method, the back propagation algorithm is chosen. For the formation of a rational training sample used in the learning of an artificial neural network, mathematical methods of temporary extrapolation are used. The analysis of the obtained results shows that the value of the mean absolute error on the test set is 3.8 %. This allows to judge the possibility of using an artificial neural network to solve the problems of predicting the SEPS state. |
topic |
forecasting of the state of the shipboard electric power system coefficient of the generalized parameter artificial neural network |
url |
http://journals.uran.ua/tarp/article/view/108528 |
work_keys_str_mv |
AT irinagvozdeva forecastingoftheperformanceoftheshipboardelectricpowersystemonthebasisoftheartificialneuralnetwork AT valerylukovtsev forecastingoftheperformanceoftheshipboardelectricpowersystemonthebasisoftheartificialneuralnetwork AT sergiitierielnyk forecastingoftheperformanceoftheshipboardelectricpowersystemonthebasisoftheartificialneuralnetwork |
_version_ |
1725063767771840512 |