In order to remedy the drawbacks of neighborhood-based crowding differential evolution for losing the roots and trapping into the local optima when solving nonlinear equations systems(NESs), this paper presents an improved differential evolution, which can be featured as follows: 1) An individual pre-judgement mechanish is proposed, which is used to divide the individuals into different groups, and different operations are used for different groups. 2) An improved hybrid differential mutation is developed to make the algorithm escape the local optima. 3) An improved archive strategy is presented to enhance the algorithm to find more roots. The experimental results on the selected test functions show that the proposed method can locate mutiple roots of the NES efficiently. Compared with other state-of-the-art methods, the propsoed method obtains better results in terms of both the root rate and the success rate.