Tasks related to the analysis of electric power network analysis are daily tasks for engineers of the power grid and fortunately, several tools are provided to carry out the specific tasks such as: load flow problem (LFP), the calculation of short-circuit analysis and stability studies. Those tools lack flexibility and sometimes precision that is needed to plan or assess the power network in real time. There is a need of an intelligent tool to help operators to take important decisions to act as fast as possible to protect the power network. Power network is a highly non-linear and complex system which is considered as a critical infrastructure for any society in the will to self-development and that should be accordingly supervised. In this paper, there is an implementation of several power network analysis tasks using intelligent agents to speed up decision processes and the precision in several algorithms is increased by replacing real numbers with two integers. A tip is used to limit up the number of digits used to approximate real numbers. The obtained results show that studied algorithms converge faster and the learning mechanism used allow to select the right method of calculating the load flow problem, build dynamic topology and compare with topologies that occurred in the past. The platform is implemented using Java Enterprise technology with Eclipse Helios, the Jademx plug-in of the Jade middleware.

component, load flow problem, fractional number, multi-agent systems, Jademx.