Switched reluctance machine(SRM) has multi - variable, strong coupling and highly nonlinear characteristics, and it s not easy to find optimal design scheme quickly and accurately by traditional methods. To solve this problem, multi-objective collaborative optimal design method was introduced. Collaborative optimization algorithm was improved, and novel dynamic relaxation factor method and elitist strategy were proposed to enhance the convergence speed and accuracy of the algorithm with guarantee of its convergence. Improved collaborative optimization algorithm was used to optimize the efficiency and torque ripple of the initial design scheme of SRM, and the global optimum values of key geometric dimensions and control parameters were obtained. According to the high power density and thermal load of the optimal scheme, a stator and rotor dual water cooling system was designed. By using lumped parameter thermal network model, the temperatures of some important parts of SRM were verified to be acceptable for practical applications. The results of analysis show that application of improved collaborative optimization algorithm can increase the efficiency of the machine and decrease its torque ripple simultaneously.