Distribution Network Monthly Line Loss Calculate and Improvement with Convolutional Neural Network

Line loss calculating is a very important problem, not only for the researchers to solve progressively, but also for power companies to reduce cost and save energy. Because of the complexity and uncertainty of distribution network structure in some countries, such as China, it gives us such a big challenge to find specific methods to calculate the theoretical line loss of a certain aera, especially by utilizing load flow calculation or other traditional methods, let alone make refine strategies and delicacy management, including power theft and leakage detection or line loss evaluation system development, etc., based on the calculation results. Based on the situation mentioned above, we create a new method by mapping the relationship of timing feature into the location of specific size picture, then use convolutional neural network to extract information from the image data to help us calculate the line loss interval and build up reasonable line loss evaluation system. Finally, a simulation model is generated and verified by using TensorFlow deep learning framework. Results fully show good accuracy and practicability of the proposed method.