Capturing the dynamic behavior of the power distribution grids, especially under high penetration of renewables, is of high interest for grid operators. The distribution power grids are not fully observable due to lack of sufficient metering infrastructure, especially downstream of medium voltage substations. Therefore, fusion of data recorded at significantly different reporting rates was proposed to increase the situational awareness of the system with non-negligible effect on the accuracy of the monitoring tool. Higher reporting rates are possible for next generation smart meters, but they raise higher concerns about data privacy, already an issue for smart meters rollout. This work proposes a framework for knowledge extraction from high reporting-rate smart metering data. The process takes place at smart meter level and with low computation and communication costs and preserving user privacy, with the scope to increase the accuracy of the monitoring tools for distribution power grids. The methodology makes use of statistical metrics able to capture system dynamics relevant for network diagnosis. The proposed approach is validated on a three-phase low voltage power flow model applied to a realistic testbed microgrid and real field measurements synchronized at one second.