DATA ADAPTATION AND TRAINING OF SIMPLE NEURAL NETWORK FOR AUTOMATION OF TELEMETRY VERIFICATION PROCESS OF THE ELECTRIC POWER INDUSTRY DISPATCHING CENTER

Abstract

The paper presents the most common construction principles for the dispatch control of the unified energy system of the Russian Federation. The approaches used in the implementation of technological tasks are described. Models of software products used to provide dispatch control, work planning and forecasting, as well as the provision of system services in the energy system of the Russian Federation, are introduced. Examples of the use of artificial neural networks are given. The specificity of information entering the dispatching centers and the difficulties that arise in the process of increasing its quantity is described. The principles of the verification of telemetry parameters and the order of their implementation established in the “System Operator of the United Power System” of the Russian Federation are noted. The concepts of software development for the verification of telemetry parameters are detailed. Algorithms and methods for preparing input data for further adaptation of the attributes of parameters affecting the final decision about truth or falsehood are presented. The principles for the further use of adapted data in calculations based on neural net-works with feedback are described. The implementation code for the simple neural network in the Python language is given. The results of experiments on two data samples are provided.