分析了ＢＰ模型学习算法——累积误差逆传播算法在接近极小点时收敛速度变得异常缓慢的原因，并通过对连接权值的调整量引入权重系数，提出了一种改进的ＢＰ模型学习算法，大大加快了收敛速度，提高了收敛性。还利用提出的改进算法对某省中期负荷进行了预测，算例结果表明了该算法的有效性。 This paper analyzes the reason why the convergence rate of the inverse model of Cumulative error propagation algorithm becomes very slow when approaching the minimum point. By introducing the weight coefficient into the adjustment of connection weights, an improved BP model learning The algorithm greatly speeds up the convergence speed and improves the convergence. The proposed improved algorithm is also used to predict the mid-term load of a province. The results show that the algorithm is effective.