Development of a platform for implementing multi-agents systems for application to automatic restoration of electric power distribution systems

• Autor(a) principal: Lucas Silveira Melo
• Data de Publicação: 2015
• Tipo de documento: Dissertação
• Idioma: por
• Título da fonte: Biblioteca Digital de Teses e Dissertações da UFC
• Texto Completo: http://www.teses.ufc.br/tde_busca/arquivo.php?codArquivo=15234
• topic: Multiagentes; RecomposiÃÃo automÃtica; Redes elÃtricas inteligentes; Multi-agents; Self-healing; Smart grids; ENGENHARIA ELETRICA; ENGENHARIA ELETRICA

It is common the occurence of permanent faults in power distribution systems. In a typical radial power distribution system when the fault protection system operates, may cause power-off not only in the fault section, but also to all customers downstream the fault.Through disjunction devices normally closed along the feeder, and normaly open on its edges, is possible to isolate the faulty sector and reenergize the healthy ones, reducing the number of customers affected by a fault. Network operators normally do this procedure manually and in addition to demand a considerable ammount of time, is subject to errors on the part of the operator. In order to automate the analisys of the network and provided it of self-healing capacity, various methods have been proposed to solve this matter. Most of these approaches adopts a centralized strategy and do not address the aspect of electric power grid self-healing. In this work is proposed an approach that uses multi-agent systems for self-healing purposes of power distribution systems. Multi-agent are highly suitable for modelling distributed systems in the smart grid domain. For a safe recovery and without violation of operational restrictions the feeder agents perform an evaluation before device agents send any command to the network switches. The proposed multi-agent system is implemented in a agentâs development platform proposed in this work that uses the Python programming language. The platform is called PADE, Python Agent DEvelpment framework. The computer representation of the network, without simplifications, is accomplished by a data encoding based on the theory of graphs and named node-depth representation that serves as a basis for the development of an API of network representation that models each of the required components in the restoration analysis. The device agents communicate with IED that in turn control the switches in the network, by means of IEC 61850 protocols: GOOSE and MMS. To validate the proposed approach, computer simulations are performed using a simplified distribution power grid as a case study and a test platform with relay test case, protection and control IED, managed switch and embedded systems.