<aside> 💡 International Journal of Applied Engineering Research ISSN 0973-4562 Volume 13, Number 21 (2018) pp. 14876-14885
</aside>
2018
Machine learning tools are computations useful in data processing to find hidden patterns or the prediction of results. The objective of this document is to compare the most used techniques of machine learning, such as: Vector Machine Support, Descriptive Discriminant Analysis, Decision Trees and Neural Networks, in Smart Grid applications. To this end, an investigation is carried out in relevant publications of the current literature.
The steady evolution of computational methods, specifically in data management and analysis has enabled several machine learning techniques to be implemented in smart grid applications. According to [1], Opower, a leading company in energy information, drives quantifiable savings through energy efficiency based on information. One of the main challenges of using smart grids is the capacity to manage electric and communication networks, where the amount of real-time information (power quality, price, energy demand, etc.) is linked to the nodes of the system. Hence, the focus lies on creating artificial intelligence models that can make decisions, even in uncertain conditions (see Figure 1) [2].
In general, definitions of artificial intelligence are related to the development of methods and algorithms that allow computers to carry out processes and make decisions in a similar way as humans do [3]. In section 2, different machine learning techniques are discussed. They can be implemented in solving problems associated with the integration and management of Smart Grids. In section 3, some applications regarding the security and efficiency of electric networks based on machine learning are shown. In section 4, a comparative evaluation of characteristics is established in order to choose an algorithm. Finally, section 5 includes a series of conclusions.
The development of computational systems has opened the door to a world of opportunities for machine learning applications through algorithms or hybrid methods that improve efficiency and are becoming progressively more powerful and capable of processing large amounts of information. The information that needs to be processed is ever-increasing, requires more accuracy, lower training times and faster response times. The migration of the electric sector towards smart grids demands the continuous development of machine learning techniques since their implementation can harmoniously integrate all the components used which grants reliability in smart electric systems as well as guaranteeing a service of quality, efficiency and continuity.