Eventos Anais de eventos
COBEM 2021
26th International Congress of Mechanical Engineering
Standards Classifier: Application to Operational Supervision of Diesel Generating Units
Submission Author:
Edgar Campos Furtado , MG , Brazil
Co-Authors:
Rafael Novais Lacerda de Oliveira, GUILHERME GOMES, Carlos Eduardo Alves da Costa, Edgar Campos Furtado
Presenter: Rafael Novais Lacerda de Oliveira
doi://10.26678/ABCM.COBEM2021.COB2021-0985
Abstract
This paper presents the development of an intelligent system for monitoring and operational supervision of Diesel-Generator Units (DGU). This system is based on computational intelligence tools, such as machine learning algorithms, focused on the preventive and predictive maintenance of the DGU. In this context, such technologies are useful tools for, in addition to other reasons, learning, detecting, diagnosing, and classifying machine operation patterns, acting assertively in assisted supervision and decision making, both at the operational level and the managerial levels. Diesel generators are an alternative source of electrical energy used to diversify the Brazilian electrical grid, aiming to mitigate events of unavailability of electricity generation and supply throughout the national territory. Essentially composed of two coupled subsystems, the mechanical and the electromagnetic, a Diesel Generator Unit becomes a system where monitoring is complex, with an extensive range of parameters from the mechanical and electrical subsystems to be correlated and supervised, to ensure the lifespan of the equipment and safety operation. Furthermore, a failure or imprecisions of supervision and operation can generate impacts on production that could affect all layers involved in the process of generation and consumption of energy, and also cause injuries to operators. The intelligent system developed in this research applies machine learning techniques inherent to the new industry 4.0 paradigm to classify operational patterns based on real generation data from more than sixty DGUs connected in the Brazilian electrical grid. To enable the learning of standards, unsupervised and supervised learning techniques were investigated. These techniques were applied to crucial parameters of the DGU, such as cylinder, water and oil temperatures, oil pressure, generator electrical current, and others, culminating in an intelligent system capable of effective assistance in the levels of control and operation of such machines. As demonstrated in this article, the developed system proved to be assertive in detecting operating incidents and optimizing the conversion of thermal to electrical power, by supporting the operating team in making decisions and adjusting critical parameters of the system.
Keywords
Diesel-Generator Unit, Intelligent system, machine learning, Operational Patterns, maintenance
