Load Flow of Radial Distribution Network by using Artificial Neural Network

Load flow analysis has an important and significant role in assessing or evaluating power system performance. There are number of traditional methods available which has higher accuracy for transmission system. In distribution networks, due to the lack of convergence, especially for large complex systems, there is need for development of alternative methods of computation. Solving equations involving Algebraic expressions are very much simpler and requires very less computer memory. There are many such literatures available that involves solving the algebraic equations alone to find the network parameters. The objective is to develop a new methodology with higher accuracy, better computation speed and simplicity. Artificial neural networks (ANNs) can be a better alternative due to their speed and ability to model nonlinear relationships. A novel technique for solving load flow problems of radial distribution networks using ANN is presented in this work. A neural network trained in back-propagation is used to map power system parameters (bus data, line data, and power demands) to phase angles and voltage magnitudes at all buses. The model is trained using data generated from conventional load flow methods. The ANN model demonstrates high accuracy, achieving negligible errors compared to conventional load flow solutions. It also exhibits significant computational advantages, giving faster solutions than traditional iterative methods. The proposed ANN-based approach provides a fast, accurate and reliable alternative for the power flow analysis of distribution networks which are mostly radial.