基于环形拓扑自适应差分进化算法优化神经网络的压气机流量特性预测

Prediction of compressor flow characteristics based on neural network optimized by ring topology adaptive differential evolution algorithm

为提升轮机模拟器中柴油机非额定工况下的仿真精度,提出一种基于有限数据的混合方法来预测多工况下的压气机流量特性.该方法采用一种基于环形拓扑差分进化算法来优化三层前馈神经网络参数,利用有限的给定工况下的流量特性数据,预测压气机流量特性曲线.采用神经网络来拟合压气机流量特性的非线性映射关系,利用差分进化算法优化调整神经网络的连接权重、连接偏置、连接开关等难以调整的重要参数.仿真结果表明,基于有限数据的方法可较好地拟合船舶压气机流量特性,采用环形拓扑的自适应差分进化算法优化调整神经网络的重要参数具有一定的有效性.本文提出的混合方法具有更好的泛化能力和搜索精度,可作为处理类似预测问题的有效手段.

In order to improve the simulation accuracy of diesel engine in marine engine room simulator under non rated conditions,a hybrid method based on limited data was proposed to predict the compressor flow characteristics under multiple conditions.In this method,a ring topology differential evolution(DE)algorithm was used to optimize the parameters of the three-layer feedforward neural network(FNN),and the flow characteristic curve of the compressor was predicted by using the limited flow characteristic data under given conditions.A FNN was used to fit the nonlinear mapping of compressor flow characteristics,and the DE algorithm was used to adjust the important parameters,such as connection weight,connection bias and connection switch.The simulation results show that the method based on limited data can better fit the flow characteristics of marine compressor,and the adaptive differential evolution algorithm with ring topology is effective to optimize and adjust the important parameters of neural network.The hybrid method proposed in this paper has better generalization ability and search accuracy,which can be used as an effective means to deal with similar prediction problems.