An Approach Based on Complex Network to Analyse and Optimise Smart Grid

In the last decades, a lot of efforts and studies have been committed to environment protection. However, it is only during the last few years that this need has become even more compelling due to our consumerist lifestyle, which produces pollution and is based on the exploitation of limited natural resources. We have been witnessed to some important green initiatives such as recycling, reafforestation, traffic blocks and energy saving measures but none of these can be considered decisive. For this reason, lately, most of researches are focused on a topic that could actually make the difference: Smart Grids. The term Smart Grid indicates a futuristic concept for an utility grid where the conventional services are coupled with ICT resources in order to manage production and consumption pro- cess in detail, avoiding waste and service interruption. This topic has gained more and more visibility thanks to the numerous dedicate conferences organized all around the world. Various organizations, including Enel in Italy, began their own smart grid project, showing great inter- est especially for the possibility to achieve a better integration with renewable. Since a smart grid can be considered as the result of a combination between engineering and computer science competences, we can list a lot of different research lines and opened challenges. After an initial period of documentation, I decided to focus my PhD on the analysis of our regional power grid in Sardinia. We had spent the first months collecting data about energy production and con- sumption from Enel and Terna, and then we used them to build a model which became the test bench for future tests. Using a specific tool for complex analysis, has been possible to depict the main characteristics of the system and, through a centrality index, we spotted the zones considered more susceptible to failures. This approach may results peculiar to some, since the term complex networks usually refers to social, biological or, in general, networks characterized by a huge number of nodes and arcs. In these architectures, the bounds between nodes are more important rather than their topological structure that, instead, is considered fundamental in power grids. However, we eventually find out that also smart grids are suitable to be analysed through complex techniques. Indeed, according to Kirchhoff's and Ohm's laws, which govern energy distribution among the areas of the grid, we can spot some analogies of their operation through the study of the patterns and clusters that naturally exists in the model. Results ob- tained by this first step of analysis, have been used as starting point for some stress tests on the grid, in order to evaluate its level of tolerance to failures. Once spotted the weak nodes, we have used them to create several scenarios where some failures and their consequences are represented. Using an optimization algorithm we studied how the performances change under harsh conditions and how the distribution ows must be reorganized to heal grid connections, making them operational. Changes in performance are registered as the difference between costs calculated as the objective functions of scenarios. Moreover, we use other parameters such as redundancy and arcs usage to evaluate the state of the grid after each failure. In order to make the experiments as more realistic as possible, we faced the problem of choosing the best metric for transmission cost, among the ones proposed in literature. We eventually decided to use a novel metric that we called power loss defined from technical data retrieved by Terna. It can be considered original, since it is created using physical parameters of the cable beyond the classic voltages and current.