Big Data na gestão eficiente das Smart Grids. HDS:Uma Plataforma Híbrida, Dinâmica e Inteligente

Resumen

In recent years, there has been an exponential increase of information generated and made available every day. Due to rapid technological advancement (e.g., mobile devices, sensors, wireless communication, etc.) billions and billions of bytes are created every day. This phenomenon, called Big Data, is characterized by 5 Vs (i.e., Volume, Velocity, Variety, Veracity, Value) and each represents real challenges (e.g., how to collect and carry a large amount of information; how to store this information; how mining it, analyzing it and extracting knowledge; how to ensure its security and privacy; how to process it in real time, etc.). It is unanimous in the scientific community that the value to be extracted from all this information will be a factor of extreme importance for the decision making, determining the success of the most varied economic areas, as well as the resolution of numerous problems. These areas include the energy ecosystem that, for ecological, economic and political reasons, led us to rethink the way we consume and produce energy. Due to the increase in energy needs caused by technological advances, the expected depletion of non-renewable energy resources and due to the energy efficiency directives imposed by the European Union, many studies have been carried out in the area of energy resources management. The term Smart Grid has emerged in the last decades with the objective of defining an intelligent energy ecosystem, which aims not only to integrate intelligence but also to automate the extremely complex operability of all its processes. Smart grids have been the subject of major studies and investments which have resulted in significant advances. However, some challenges have to be addressed in the management of its complex data flow. It is in this context that the present dissertation falls, with the main objective on obtaining solutions to some of the problems identified in the field of Smart Grids using new techniques and methodologies proposed in the area of Big Data. This paper presents a study on the recent and growing technological advances in the area of Big Data, where its major challenges are identified. These include complexity in the management of continuous and disordered flows, the need to reduce the time spent in pre-preparation of data and the challenge of exploring solutions that provide analytical automation. On the other hand, the study analyzes the impact of the application in the new technologies in the development of the Smart Grids, in which it is concluded that, although embryonic, its application is essential for the evolution of the energy ecosystem. This study also resulted in the identification of the main challenges in the area of Smart Grids, which highlight the complexity in managing its data flow in real time and the need to improve the accuracy of energy consumption and production forecasts. Given the identified challenges, a conceptual model, based on the Docker Container architecture, was proposed for the development of a platform. This model aims at flexibility and agility in order to allow the integration and validation of the new and growing technological approaches proposed in the area of Big Data, necessary for the development of Smart Grids. In order to validate the proposed model, a stack was developed where several services were implemented that aimed to contribute to the challenges identified in the area of Big Data and Smart Grids, namely: visualization and monitoring of data collected in real time; preparation of data collected in real time; real-time forecasting of multiple time series simultaneously; detection of anomalies; evaluation of the accuracy of forecasting and generation of new models for the forecast of consumption and production of energy according to certain criteria. Finally, a number of case studies were developed whose results allowed us to conclude on the importance of the pre-preparation of the data in the analytical phase, on the efficiency in the analytical automation and on the advantages of the Edge Analytics. Unlike more traditional approaches to the centralized execution of the analytic process, edge analytics explores the possibility of performing data analysis in a decentralized way from a non-central point of the system. The results allowed to conclude that edge analytics brings added advantages to the precision of the forecasts. Results allowed us to infer how to collect the data in order to obtain a better precision in the predictions, i.e., the more precise and context-adjusted the forecasts are executed the greater their accuracy.