•The multi-island genetic algorithm (MIGA) is applied to optimize the configuration of a universal selective solar absorber, which is superior to traditional genetic algorithm (GA) in finding the global optimum.•The optimization design is conducted to achieve maximum absorptance A(λ) over solar spectrum and minimum emittance ε(λ) in the mid-infrared band at the same time.•A set of optimal geometric parameters are obtained, with the solar absorptance greater than 0.932 with 96% probability and mid-infrared emittance lower than 0.058 simultaneously.•The optimal absorber is a combination of great characteristics, such as highly spectral selectivity, polarization independence and angle insensitivity. Based on the demand for ideal selective solar absorbers to harvest a full solar spectrum, we present an auspicious method, the multi-island genetic algorithm (MIGA), to optimize the configuration of a universal selective solar absorber, which can avoid the premature phenomenon and superior to traditional genetic algorithm. The selective solar absorber equipped with the set of optimal geometric parameters exhibits the solar absorptance greater than 0.932 with 96% probability and mid-infrared emittance lower than 0.058 at the same time considering unavoidable 20% fabrication uncertainties in the design variables. In addition, the absorber is a combination of great characteristics, such as polarization independence and angle insensitivity. The hybrid resonance modes of Wood’s anomaly, surface plasmon polariton and magnetic polariton are responsible for the high absorption performance. The MIGA method presented is proved to be an effective and robust tool in the optimization of micro-nano structures with manufacturability for thermal and energy applications.