Visibility of Power Distribution Network Using Neural Network

Visibility and analysis of power networks and distribution of electricity are among the basic needs of electricity companies. Conventional methods of load calculation and distribution are very time consuming and complex. In this article, using neural network and machine learning algorithms, the voltage of different parts of the electrical network is investigated. In the proposed method, there is no need for different network parameters and even how the network is arranged because all the involved parameters are trained by neural network neurons. Finally, by simulation and machine learning, the unknown voltage of different parts of the network is obtained, through which the entire network is studied.

I. Introduction

Load distribution analysis is the main basis of power system analysis, which is necessary for planning, operation or economic planning and exchange of electricity as well as utility services [1]. Load distribution is one of the most valuable tasks that is the concern of operating engineers and supervisors of systems and power companies. Its main purpose is to calculate all the values of voltage and current and current angles of the lines and thus the load distribution through transmission lines or distribution networks. Stable operating conditions of power systems are determined by performing load distribution analysis on subsystems [2]. Many methods are used to perform load distribution calculations, the most famous of which are Gauss Seidel and Newton Raphson. These methods are due to unbalanced network, radial structure or with weak network, many branches and nodes, high resistance to reactance ratio, wide resistance and reactance, wide range of resistance and reactance, multiplicity of points, lack of measuring equipment for all Branch current measurements lose their flexibility and performance in distribution networks and are not suitable for use in these types of networks [3]. Another method that is widely used in power grids is the backward-forward method. This method was first presented by Berg et al [4]. The principles of operation of this method are based on repetition, each repetition consists of two levels: the first stage, in which the current of all branches of the network is calculated (inverse level) and the second level, in which the voltage by the branches and the impedance of the branches, the bus voltage It is calculated (advancing step) and with the values obtained from the bus voltage, the bus current (load) values are updated. The above methods are valid for the steady state of the network.