Abstract: Power forecast of gas turbine is significant for breakdown prediction. For gas turbine failure prediction, this paper proposes a power prediction method based on Elman neural network. Experimental data are from the real data of a power plant, this paper selects the attributes relevant to power. Through comparative experiments, using suitable preprocessing method and input data, and setting the suitable number of neurons in hidden layer, this paper develops a gas turbine power prediction model based on Elman neural network. Finally, compared with back propagation (BP) network and radial basis function (RBF) network, this paper verifies the effectiveness of Elman neural network.

Key words: Elman neural network, gas turbine power prediction, neural network

摘要： 预测燃气轮机未来的功率变化趋势对故障预测非常重要。针对燃气轮机故障预测的问题，提出了一种基于Elman神经网络的功率预测方法。以某电厂燃气轮机的实际数据为例，选取与功率变化最相关的属性。通过对比实验，采取合适的预处理方法，确定神经网络模型的输入，设置合适的隐含层神经元个数，从而建立了基于Elman神经网络的燃气轮机功率预测模型。最后通过与反向传播（back propagation，BP）网络、径向基函数（radial basis function，RBF）网络进行比较，验证了该方法的有效性。

关键词: Elman神经网络, 燃气轮机功率预测, 神经网络