In this paper, the authors propose a new protection strategy for electrical Medium Voltage distribution networks, based on a distributed intelligence system composed of a central unit and of many local processing units. The distributed intelligence system can precisely identify and locate faults in suitably limited times. Each local processing unit is composed of a data acquisition system and a data analysis system. The first allows the accurate reconstruction of the fundamental using only 1.5 cycles of under fault transient electrical quantities. The second performs suitably precise on-line fault diagnosis based on the data provided by the data acquisition module. The data acquisition algorithm and the data analysis algorithm are managed by the data processing system of the remote processing units implemented on a micro-processor. With the proposed architecture, if local acquisition, elaboration and clearing times are limited, the faulted line experiences only two micro-interruptions, also during multi-phase permanent faults, with a great improvement of the quality of the service. Some experimental results on a simulated test system, proving the efficacy and simplicity of the proposed approach, are reported.