基于神经网络的电磁干扰的预测

提出了一种应用神经网络预测电磁干扰的方法.针对遗传算法总体搜索能力较强但容易陷入局部最优,而模拟退火算法具有较强的局部搜索能力,又能避免搜索陷入局部最优解的特点,将模拟退火算法与遗传算法相结合,优化多层前馈(BP, Back Propagation)神经网络,获取最优的权值和阈值,并采用模拟退火的思想确定隐含层神经元的个数,进而建立基于神经网络的电磁干扰预测模型.以双平行导线间的电磁干扰问题为实例,明确干扰要素,建立训练样本和测试样本,对比期望输出和预测输出之间的误差,结果表明该方法可以准确有效地进行电磁干扰预测.

Prediction of electromagnetic interference based on neural network

A method to predict the electromagnetic interference using neural network was proposed. Genetic algorithm has the strong overall search ability but easy to fall into local optimum, and simulated annealing algorithm has the partial search ability, avoiding the search into local optimal solution. By using the simulated annealing algorithm and genetic algorithm combining, the back propagation (BP) neural network weights and thresholds were optimized, and the number of hidden layer neurons was determined by the simulated annealing ideas. Then, the neural network-based predictive models of electromagnetic interference was established. With the two parallel leads to electromagnetic interference matter as predicted instance, interference factors were identified, and the training and test samples were established. In contrast to the error between the expected output and the predicted output, the results show that the method can accurately predict the electromagnetic interference effectively.