The increasing penetration level of photovoltaic (PV) power generation in low voltage (LV) networks results in voltage rise issues, particularly at the end of the feeders. In order to mitigate this problem, several strategies, such as grid reinforcement, transformer tap change, demand-side management, active power curtailment, and reactive power optimization methods, show their contribution to voltage support, yet still limited. This paper proposes a coordinated volt-var control architecture between the LV distribution transformer and solar inverters to optimize the PV power penetration level in a representative LV network in Bornholm Island using a multi-objective genetic algorithm. The approach is to increase the reactive power contribution of the inverters closest to the transformer during overvoltage conditions. Two standard reactive power control concepts, cosφ(P) and Q(U), are simulated and compared in terms of network power losses and voltage level along the feeder. As a practical implementation, a reconfigurable hardware is used for developing a testing platform based on real-time measurements to regulate the reactive power level. The proposed testing platform has been developed within PVNET.dk project, which targets to study the approaches for large PV power integration into the network, without the need of reinforcement.