Confirmation of Electrical Network Configuration from Telemetry Data Based on Convolutional Neural Networks

The reliability function of electrical-network configuration is an important part of decision-support systems and increases the reliability and efficiency of managing the operating mode of an electrical network. Classically, the function is implemented on the basis of the condition assessment method and does not have the proper response rate to changes in the electrical network, and also requires large computing resources. Meanwhile, there are modern tools that ensure a high reaction rate due to the exclusion of calculations of the electrical network mode in real-time control of it. Such tools include neural networks. The article proposes a system for confirming electrical network configurations based on convolutional neural networks. The initial data for the neural network are telemetry data of the operating parameters presented in 2D format, and the result is localized errors in the representation of the current configuration of the electrical network with an assessment of the level of reliability of the result. The operation of the system is demonstrated by the example of evaluating the configurations of a nine-node electrical circuit of the IEEE standard. The neural network demonstrates high efficiency and accuracy of recognition of the current configuration, including in conditions of distortion and insufficiency of telemetry.