The quality of injected current in multi-paralleled grid-connected inverters is a matter of concern. The current controlled grid-connected inverters with LCL filter are widely used in the distributed generation (DG) systems due to their fast dynamic response and better power features. However, designing a reliable control system for grid-connected inverters with LCL filter is complicated. Firstly, overcoming to system resonances due to LCL filters is a challenging task, intrinsically. This could become worse as number of paralleled grid-connected inverters increased. In order to deal with resonances in the system, damping methods such as passive or active damping is necessary. Secondly and perhaps more importantly, paralleled grid-connected inverters in a microgrid are coupled due to grid impedance. Generally, the coupling effect is not taken into account when designing the control systems. In consequence, depending on the grid impedance and the number of paralleled inverters, the inverters installed in a microgrid do not behave as expected. In other words, with a proper control system, a single inverter is stable in grid-connected system, but goes toward instability with parallel connection of other inverters. Therefore, consideration of coupling effect in the multi-paralleled grid-connected inverters is vital. Designing control systems for multi-paralleled grid-connected inverters becomes much more difficult when the inverters have different characteristics such as LCL filters and rated powers. In this paper, the inverters with different characteristics in a microgrid are modeled as a multivariable system. The comprehensive analysis is carried out and the coupling effect is described. Also, the control system design for multi-paralleled grid-connected inverters with LCL filter is clarified and a dual-loop active damping control with capacitor current feedback is designed. Finally, the proposed multivariable control system for a microgrid with three-paralleled grid-connected inverters with LCL filter is validated by simulation.