Estimation of load curves in electrical transformers by means of neural networks

Load curves in electrical transformers are employed to quantify losses in power distribution systems, to establish power utility rates that vary according to the hour of the day in which energy consumption occurs, and to optimize the use of transformers. This paper presents a method to estimate load curves in distribution transformers. From measurements made on a sample of transformers and the available information from the commercial database of an electrical power distribution company, a neural network was trained by means of a MATLAB-based computational tool. The load curve estimation method was implemented in the power distribution system of the city of Buenaventura, Colombia. The neural network learning and the load-curve estimation method were validated through their application to a set of transformers different from those used for training the neural network, obtaining an accuracy higher than 90 % for the load curve estimation. As compared to traditional methods, the method presented in this work allows to estimate the load curve of any distribution transformer in a more economical and accurate way.