To address the issue of high implementation complexity in compensating for nonlinear distortion in short-range optical fiber communication systems using Intensity Modulation and Direct Detection (IMDD), a method based on an Adaptive Artificial Neural Network (ANN) equalizer is proposed. By considering the characteristics of optical communication networks, the number of nodes in the input layer and hidden layer of the ANN, as well as the number of training samples, are determined. Taking into account the noise in the optical fiber system and the distortion caused by dispersion, the training samples are expanded to enhance the generalization capability of the ANN equalizer. The weights of the ANN equalizer are adaptively adjusted, and small changes in the weight values are utilized to track channel fluctuations, thereby alleviating the problem of weight offset caused by variations in the fiber channel parameters due to changes in environmental conditions. Experimental results show that compared to a non-adaptive ANN, the proposed adaptive ANN has advantages in terms of overall gain, computational complexity, and memory requirements. The model demonstrates strong robustness in optical fiber communication under noisy conditions and has practical value.