本文从一个新的角度研究了基于人工神经网络 (ANN)的预测问题 ,针对在线参数的特点 ,提出一种基于ANN的自适应预测算法 ,研究了训练样本噪声 ,采样频率与信号频率对算法性能的影响 ,对算法作出改进。并从预测算法的通用性和速度等方面对该算法进行了评价。将该算法应用于电力设备在线参数的预报 ,预测结果表明该算法是有效的 In this paper, a new ANN-based prediction problem is studied. According to the characteristics of online parameters, an ANN-based adaptive prediction algorithm is proposed. The effects of training samples noise, sampling frequency and signal frequency on performance of the algorithm The impact of the algorithm to make improvements. The algorithm is evaluated from the aspects of universality and speed of prediction algorithm. The algorithm is applied to predict the online parameters of power equipment. The prediction results show that the algorithm is effective