Application of a Risk-Based Approach and Deep Convolutional Neural Networks to Determine the Set of Flight Points in the Diagnostics and Design of Electrical Facilities

The article deals with the problem of determining suboptimal spatial points and angles for diagnosing the technical state of elements of electrical facilities based on multispectral photography from an unmanned aircraft. As an object of analysis, a power plant with tube-mounted overhead connections is used. An important distinctive feature of data collection using unmanned aerial vehicles on the territory of critical infrastructure facility is the need to maximize flight safety. The article proposes a new formulation of the problem of using unmanned aerial vehicle for diagnosing electrical systems, taking into account the economic effect of data collection, the risks of both equipment failures and the probable accidents of an unmanned aircraft. At the same time, it is required to use a digital twin (digital instance) to assess the solutions generated by the neural network, since it is necessary to model how the diagnosed equipment elements will be viewed from the unmanned aerial vehicle route points, have a spatial traffic safety map at the substation and a map of all diagnosed objects with additional data. The developed technique can also be used in the design of new electrical facilities in order to increase the efficiency of their diagnostic procedures during subsequent operation.