The ORC system of a diesel engine with the non-azeotropic mixture was investigated and  a highdimensional evolutionary multi-objective optimization(EMO) method was proposed based on balanced and unbalanced weights. The results indicate that using non-azeotropic mixtures in the ORC system enhances the match between the working fluid and heat source in the heat exchanger. Compared with the pure working fluid, the output power(W_net) of benzene/toluene (60%/40% mass ratio) increases by 5.4%-19.2%. According to the sensitivity parameter analysis, W_net first increases and then decreases, the thermal efficiency(η_th) monotonically increases, and the total exergy destruction(I_zon) monotonically decreases when the evaporation temperature increases. The W_net, η_th, and I_zon indicators are mutually exclusive, and so there is no operating parameter that satisfies all three indicators simultaneously. Consequently, a high-dimensional EMO method is employed, and multi-objective optimization is performed using RPD-NSGA-Ⅱ and analysis is conducted by TOPSIS. Based on the balanced weight approach, the optimal values forη_th, W_net, and I_zon are 24.07%, 72.36 kW and 44.66 kW respectively with a diesel exhaust gas energy recovery rate(η_g) of 12.5%. On the other hand, choosing unbalanced weights of W_net andη_th results in the optimalη_th and W_net of 22.42% and 81.44 kW, which corresponds to the I_zon andη_g of 52.33 kW and 14.07%. Accordingly, the use of different weight combinations presents more engineering choices.