In face of the conventional attribute reduction incapability of grid data preprocessing for its big volume

diversified types and high speed in the forthcoming age of big data

a new method of electric power big data preprocessing via attribute reduction based on cloud computing technology is put forward. The characteristics of the relative positive region theory of the rough set is analyzed

and a parallel attribute reducing algorithm named MP_POSRS that can calculate the number of elements in the relative positive region is designed by taking good advantages of the MapReduce model in this method. Finally

the experiments including operations on the decision table of power grid fault diagnosis and real data of wind power are performed on a Hadoop platform

the results showing that the method is effective and feasible for dealing with power grid big data

and with good speedup and scalability required by electric power big data preprocessing via attribute reduction.