The capacity-to-load ratio is an important indicator to guide the power grid planning, and has a decisive influence on the selection of substation capacities and the construction of power lines. A capacity-to-load ratio optimization method considering collaborative planning of multi-voltage level power grids is proposed in this paper, which aims to minimize the construction cost and takes into account the synergetic planning of substations and power lines of multi-voltage levels. By combining the Bayesian optimization algorithm (BOA) and genetic algorithms (GA), a BOA-GA solving algorithm for the planning model is proposed. This algorithm automatically generates planning schemes for multi-voltage level power grids, and obtains the optimal capacity-to-load ratio, as well as the optimal investment of the power grid construction. Additionally, the study reveals that the relationship between capacity-to-load ratio and the overall investment of power grid construction is a quadratic function over a long time scale, and the curve is a parabola going upwards. Case studies demonstrate that the proposed model can achieve optimal configuration of multi-voltage level capacity-to-load ratio and automatically plan the network structure. The construction cost is significantly reduced while various technical indicators are satisfied. The proposed method can provide effective theoretical and methodological guidance for multi-voltage level power grid synergetic planning.