Microchannel heat sink has the advantages of high heat transfer efficiency and reliability, and it is useful for cooling micro-scale high heat flux density electronic components. In order to meet the performance requirements and control cost, it is necessary to optimize the heat transfer capability and flow resistance of the microchannel heat sink at the same time. The thermal resistance network model used in the traditional research is relatively simple, which can not well reflect the response of thermal resistance and flow resistance to the topological changes of the microchannel cross-section shape, and the optimization object is usually the size of a given shape section. A single-layer silicon substrate microchannel heat sink thermal resistance network model was proposed by discretization method. The heat sink fins were separated into smaller ones. According to the response of microthermal resistance to the width of microdots, and the contribution of microthermal resistance to the overall thermal resistance, the overall thermal resistance of microchannel was described. Taking the output power of micro-pump as the optimized boundary condition and the pressure drop and thermal resistance as the optimization targets, the size of quadrilateral uniform cross-section silicon substrate heat sink was optimized by SQP (sequential quadratic programming) method. The optimization results were simulated and verified by CFD (computational fluid dynamics). The results showed that the shape of the cross section was rectangular when the height of fin was low, and it gradually turned into a triangle as the height of the fin increased. In the design range, when the microchannel section was trapezoidal, the fin section was triangular, the heat transfer efficiency and pressure drop were ralatively dominant. Under boundary point method and ideal point method, the optimization results of microchannel height, fin width, groove bottom width and groove top width were 500, 50, 64.5, 114.5 μm and 500, 50, 50, 100 μm, respectively. This method can adjust the evaluation function according to the design requirements, meanwhile, the calculation result has important engineering significance and provides reference for designers.