Industrial Automatic Control Systems and Controllers Annotation << Back Neural Network Method of Monitoring the Operating Modes of Substation Information-measuring Systems for Electricity Metering A.A. Soldatov, Yu.K. Evdokimov The peculiarities of using the volt-ampere parameters of the substation electric power metering devices in determining the state of the operation mode of the substation information-measuring complex of electricity metering and their use as input attributes (parameters) for the classifi cation of the state’s «norm», «fail» and «not determined» are shown. The method of neural network control of operating modes of substation information-measuring complexes for electricity metering is proposed. The selection of the type of artifi cial neural network (INS), its architecture, the number of neurons at the input of the network, in the inner layer and at the output, based on ensuring the greatest reliability of the network classifi cation is substantiated. The method of network training and the application of the hyperbolic tangent activation function are shown. A quantitative analysis of the structure of the parameters of the trained neural network was carried out to classify the state of operation of the substation IIC, which showed that multilayer perceptron type networks produce better results than RBF neurons in solving the problem under consideration. The advantage of application of the neural network forecasting method in solving problems of verifi cation of the state of operating modes of substation information-measuring complexes for electricity metering in comparison with the classical statistical method for the normal law of normal distribution of input parameters (attributes) is proved. Quantitatively this advantage is expressed in the increased sensitivity of the ANS method at 34 % and reliability, which is 1.4 times better. Keywords: neural network control; modes of operation of information and measurement systems for electricity metering. Contacts: E-mail: aa.soldatov@bk.ru, E-mail: evdokimov1@mail.ru Pp. 35-49.