基于蚁群算法的神经网络绝缘子闪络电压预测

为提高覆冰绝缘子闪络电压的预测精度及预测速度,采用BP神经网络和蚁群算法相结合的方法进行预测模型设计。利用闪络电压及其影响因素之间的试验数据,建立其神经网络的预测模型。以网络的权值和阈值为自变量,通过蚁群算法的迭代运算,搜索出误差全局最小值,再进行网络的二次学习训练。结果表明,该方法具有较高的预测精度,适用于绝缘子闪络电压的预测。

Insulator Flashover Voltage Forecasting Based on Neural Network and Ant Colony Algorithm

To improve prediction precision and efficiency of icing insulator flashover voltage, combination of BP neural network with ant colony algorithm is proposed to design prediction model. According to the experimental data of flashover voltage and its impact factors, the neural network prediction model is established. Taking neural network weights and threshold values as decision variables, the global minimum error can be obtained by using iterative computation with ant colony. Then secondary training of neural network can be accomplished. Example results show that the method has high prediction accuracy and it is suitable for prediction of insulator flashover voltage.