The development of new digital traction power system is very rapid. And a large amount of heterogeneous data will be generated from the distribution network edge terminal device under the infiltration of renewable new energy. It will bring great data load pressure to the distribution network operation and maintenance platform. This problem causes data processing delay and makes user-side response unreal-time. For solving this problem, deep data mining and weight removal on the distribution side are needed to reduce. Traditional mining techniques for heterogeneous data generally use data mining methods based on dynamic time regularity. But the disadvantage of this method is that the mining efficiency is too low. It can not be weighted for multi-source heterogeneous data with low similarity. For this reason, this paper presents an edge-side processing algorithm for heterogeneous data based on Euclidean distance weighted optimization, which calculates the similarity between heterogeneous data from multiple sources to eliminate data redundancy, data mining and weighting. Finally, the effect of data mining technology in the edge computing security protection system is analyzed and verified based on the actual sample data of the distribution network comprehensive energy station.