基于自适应RBF神经网络的变压器噪声有源控制算法

针对现有变压器噪声有源控制算法存在的不足,提出了一种用于抑制噪声的新算法。该算法融合了自适应算法、粒子群算法、改进梯度下降算法及RBF神经网络算法。首先利用自适应算法确定降噪系统控制器中RBF神经网络隐含层节点个数和相应的参数;然后,根据切换策略自适应地选择粒子群算法或者改进梯度下降算法,用来优化节点数目和参数;最后,将优化得到的隐含层结构和参数反馈至系统控制器中,使系统的次级声源更好地抵消源声源。通过将所提的改进RBF神经网络法与未改进的RBF神经网络法和BP神经网络法进行比较,表明该算法可有效地提高降噪系统的自适应能力和抗干扰能力,且能够将噪声控制在较低的范围内,获得较理想的降噪效果。

Novel Algorithm of Active Control for Transformer Noise based on Adaptive RBF Neural Network

For the deficiency of the existing active noise control algorithms for transformers, a new algorithm for noise suppression is proposed, which is based on the combination of several algorithms, i.e. adaptive algorithm, particle swarm optimization, improved gradient descent algorithm and RBF neural network algorithm. Firstly, the algorithm applies the adaptive algorithm to determine the number of nodes and the corresponding parameters of the hidden layer of RBF neural network in system controller; Then, according to the switching strategy, particle swarm optimization or improved gradient descent algorithm are selected adaptively to optimize the node number and parameters; Finally, the optimized nodes and parameters of the hidden layer are fed back to the system controller. As a result, the infrasound source of the system is able to better offset the initial sound source. By comparing the improved RBF neural network proposed with the conventional RBF neural network and BP neural network, it is shown that the proposed algorithm can effectively improve the adaptive ability and the anti-interference capability of the system, by which the transformer noise can be controlled within the relatively low range and the improved ability of noise reduction is obtained.