Development of a Deep Learning based attack detection system for Smart Grids

Abstract

The electrical power grid is evolving into a more modern electric grid, called smart grid. Smart grids use two-way flows of electricity and information to create an automated and distributed advanced energy delivery network. This improvement has required the addition of a large number of new elements, which contribute to the appearance of new vulnerabilities and therefore, possible new attacks. Thus, security has become one of the most important issues in terms of distribution and generation of energy. For this reason, there is a need to detect these attacks and thus, reduce the loss of money that they cause. Consequently, the objective of this project is to develop a system that automatically detects these attacks using Deep Learning techniques. For that purpose, the following tasks will be carried out: First, it is necessary to study the current state of the art of different areas: the types of attacks that smart grids can suffer, different tools that will allow us to obtain smart grids data and the deep learning techniques and tools that will be necessary. As it is not possible to obtain real data from smart grids because it is confidential, Agent-Based Model Simulation will be used in order to simulate them. In this simulation different attacks studied previously will take place and synthetic data will be generated. Once the data is obtained, it will be processed and analyzed using big data techniques. In addition, deep learning algorithms will be applied to them in order to make inferences about the behavior of the grid. Finally, the data obtained from the neural network will be processed in order to detect the attacks. This system provides the possibility of automating the process of detecting attacks on smart grids, avoiding the great loss of money to companies caused by electrical fraud.