|
|
THEME OF MASTER'S WORK:
"Neuronetwork modelling of tasks of electric networks operative management"
Actuality of theme.
Actuality to perfect the management of electroenergy objects on the modern stage of development of energy increases because of the change of their operating conditions. To provide the proper level of reliability of functioning of electroenergy objects there is an insistent necessity to perfect the control, especially the dispatcher`s control by development of new approaches, including, ones based on the methods of artificial intelligence and principles of adaptive control.
Purpose of researches.
The purpose of this work is to find out the progress of operative management trends on the basis of intellectual methods with an orientation on existent hardwares taking into account their perspective development and on adaptation of model constituents to the features of concrete technological task and management purpose.
Presenting of basic material.
As it follows from a large number of works [1-15 and other], mainly by foreign authors, three types of Artificial Neural Network (ANN) are commonly used: multi-layered networks of direct distribution, networks of Kokhonena, recurrent networks of Khopfil'da. The features of neuronetwork modelling initiated the great number of researches of its application while solving different tasks of operative management of EES, that is most completely reflected in [16].
The important stage in creation of ANN is its training which consists in adjusting the parameters of ANN. The type of ANN determines training features.
Training of multi-layered ANN.
There are optimum values of outputs of neurons of all layers in networks, except for the last, one as a rule, not known. Under such conditions it is impossible to train multi-layered pertseptron taring into account only the sizes of errors on the outputs of ANN. Teaching of multi-layered networks is supervisory, requiring the selection of presence of not only great number of vectors of entrances but also great number of the proper responses.
Mathematically a task consists of finding such values of gravimetric coefficients (at the fixed structure), that the error of disagreement between the reaction of network and by the required response for all examples of teaching selection was minimized. Adding up is made on all neurons of an output layer and on all images processed by a network:
 (1)
where yij, dij - accordingly is actual and desired reaction of j neuron of an output layer on an I entrance vector on, p is a number of images in a teaching selection, m is a number of neurons in an output layer.
Information, which is entered into ANN in the process of teaching, must be kept in the interneuron connections called - sinapsis. Thus, teaching consists of modification of sinaptical scales of neurons.
Conclusions.
1. I have analysed the use of ANN for operative management of the electroenergetic systems.
2. Directions how to increase the fast-action of the control system and problem of their construction are shown.
Now master's degree work is on the stage of development.
Hypothetical time of completion: January, 2008.
Literature
1 Peng T.M., Hubele N.F., Karady G.G.. Conceptual approach to the application neural networks for short-term load forecasting // IEEE Int. Symp. Circuits and Syst. - 1990. - vol.4. - Р. 2942-2945.
2. Kottathra K.,.Pryor T.L, Cole G.R., Cheok K. Load Forecasting for remote area power supply systems. //The 11th Conference on Artificial Intelligence for Applications (Car.No.95CH35758), Los Angeles, CA, USA, 20-23 Feb. 1995 (Los Alamitos, CA, USA: IEEE Comput. Soc. Press 1995) - Р. 231.
3. Bakirtzis A.G. and oth. Short term load Forecasting Using Fuzzy neural networks // IEEE Power Eng. Review. 1995. - vol.10. - N3.
4. Kwang-Ho Kim. Implementation of Hybrid short-term load Forecasting System Using Artificial Neural Networks and Fuzzy Expert systems // IEEE Power Eng. Review. – 1995. - vol.10. - N3.
5. Bakirtzis A.G. and oth. A neural network short term load forecasting model for the Greek Power system // IEEE Trans. on Power Systems. – 1996. - vol.11. - N2.
6. Simo J.B., Kamwa I.. Exploratory Assessment of the Dynamic Behavior of Multimachine System Stabilized by a SMES Unit // IEEE Power Eng. - 1995.
7. King T.D. and oth. Optimal environmental dispatching of electric power systems via an improved Hopfield neural network model // IEEE Trans on Power Systems. – 1995. - v.10. - N3.
8. Sidhu, T.S., Ao Z. On-line Evaluation of capacity and Energy Losses on Power Transmission Systems by using Artificial neural networks // IEEE Trans. on Power Delivery. -1995. - v.10. - N4.
9. Bernieri A., Betta G.. Neural network and psenvo-measurements for real-time monitoring of distribution systems // IEEE Trans. on Instrumentation and measurement. – 1996. - v.45. - N2.
10. Park Y.M. and oth. A neural-network based Power system stabilizer ising Power Flow Characteristics // IEEE Trans. on Energy Conversion. – 1996. - v.11. - N2.
11. Hayashi Y. and oth. Efficient Determination of optimal Radial power system structure using Hopfield neural network with Constrained noise // IEEE Trans. on Power Delivery.-1996.-v.11-N3.
12. Fushuan W., Zhenxiang H. A kind of generalized Hopfield continuons model and its application to the optimal distribution of reactive sources in power systems. Artificial neural network (ed. T.Kohonen) 1991.
13. Hiyama T. and oth. Evaluation of neural networks based Re-al time maximum power tracking Controller for PV system // IEEE Trans. on Energy Conversion. – 1995. - v.10. - N3.
14. Чукреев Ю.Я., Хохлов М.В., Готман Н.Э. Применение искусственных нейронных сетей в задачах оперативного управления режимами электроэнергетических систем - Сыктывкар: Коми НЦ УрО РАН. - Вып. 56, 2000. 24 с.
15. Чукреев Ю.Я., Хохлов М.В., Алла Э.А. Оперативное управление режимами региональной энергосистемы с использованием технологии искусственных нейронных сетей //Электричество. - 2000. - №4. - С. 2-10.
16. Neubor D.et al. Artificial neural networks for power systems/ CIGRE TF 38.06.06 // Electra. - 1995.- №159. - P. 77-101.
|
|
|
|
|
