基于GA-RBF神经网络的变压器温升预测

以预测变压器温升为目的,提出了一种基于遗传算法(GA-Genetic Algorithm)优化径向基函数(RBF-Radial Basis Function)神经网络的预测模型。首先用GA算法优化RBF神经网络中的隐层节点个数、输出权重、隐层基函数中心及宽度这四个参数的初值,然后利用优化后的RBF神经网络对样本进行训练,这样克服了传统神经网络参数选择的随机性。以S9-1000/10低损耗电力变压器为例作温升试验,将预测值与实测值对比,并与基于传统的BP神经网络预测值对比,结果表明,该方法得到的变压器温升预测值与实测值更接近,该预测模型具有更高的精度和适应能力。

Transformer Temperature Rising Forecasting Based on GA-RBF Neural Network

A genetic algorithm optimization of radial basis function neural network prediction model for the prediction of transformer temperature rising is presented.Firstly,it uses the GA algorithm to optimize the RBF neural network initial value of four parameters including the number of hidden layer nodes,the output weights,the hidden layer basis function centers and width,then it uses the optimized RBF neural network to train samples,which overcomes the random parameters of the traditional neural network.Taking an S9-1000/10 low-loss power transformer as an example for the temperature rising test,the predicted values are compared with measured values and the values based on traditional BP neural network prediction.The results show that transformer temperature rise predicted values using this method is closer to measured values,and this prediction model has better accuracy and adaptability.