As the construction of new power systems advances，the spatio-temporal mismatch between the output of intermittent renewable energy and diversified load exacerbates the peak-valley difference in power grids，leading to an increased load rate of main transformers and a decline in load balance of each transformer within substations.This operational trend accelerates the risk of equipment aging and poses potential threats to the safe and stable operation of the power grid through hierarchical coupling effects.To address this issue，based on the hierarchical partition operation characteristics of substations and the spatio-temporal coupling relationship of outgoing line loads，a multidimensional evaluation index of inter-line correlation is proposed，which incorporates load similarity and complementarity.By comprehensively taking into account the load rate，load flatness，and load balance of the main transformers，a three-dimensional criterion for optimal load distribution of main transformers is proposed.Furthermore，a multi-objective optimization model considering load correlation characteristics is developed，and the solution methods of the enumeration method and genetic algorithm are employed to obtain the Pareto-optimal solution set under multiple constraints.Empirical validation analysis based on typical 110/35 kV substation cases in East China demonstrates that the proposed method reduces the standard deviation of the daily load rate of main transformers by 42.7% and the peak-valley difference rate by 18.3%，effectively enhancing equipment utilization efficiency and operational economy，which provides an innovative solution for the lean operation and maintenance of intelligent substations.