Via the integration of renewable energy and storage technologies, buildings have started to change from passive (electricity) consumers to active prosumer microgrids. Along with this development come a shift from centralized to distributed production and consumption models as well as discussions about the introduction of variable demand-supply-driven grid electricity prices. Together with upcoming ICT and automation technologies, these developments open space to a wide range of novel energy management and energy trading possibilities to optimally use available energy resources. However, what is considered as an optimal energy management and trading strategy heavily depends on the individual objectives and needs of a microgrid operator. Accordingly, elaborating the most suitable strategy for each particular system configuration and operator need can become quite a complex and time-consuming task, which can massively benefit from computational support. In this article, we introduce a bio-inspired cognitive decision agent architecture for optimized, goal-specific energy management in (interconnected) microgrids, which are additionally connected to the main electricity grid. For evaluating the performance of the architecture, a number of test cases are specified targeting objectives like local photovoltaic energy consumption maximization and financial gain maximization. Obtained outcomes are compared against a modified simulating annealing optimization approach in terms of objective achievement and computational effort. Results demonstrate that the cognitive decision agent architecture yields improved optimization results in comparison to the state of the art reference method while at the same time requiring significantly less computation time. (C) 2014 Elsevier Ltd. All rights reserved.