基于线性神经网络的煤层气发动机空燃比动态建模

Dynamical Modeling Based on the Linear Neural Networks for the Coal-bed Gas Engine Air Fuel Ratio

针对预混合双阀控制的煤层气发动机,利用引入动态环节的递归线性神经网络,基于台架实验所获得的数据建立煤层气发动机空燃比动态过程的输入/输出模型。以均方差为评价指标确定模型输入、输出的阶次,用非建模数据对训练后的模型进行验证。结果表明,模型准确复现了不同工况下空气和燃气质量流量与排气空燃比的动态特性。

Aimed at a pre-mixed coal-bed gas engine with double valves control system, based on the experimental data gained on the engine test-bed, a dynamic input/output model for the coal-bed gas engine air fuel ratio control was built using the recursion linear neural networks with introduced dynamic loop. The input/out- put model orders were determined by mean square errors （MSE） as evaluating index, and the trained model was verified using non-modeling data. The results show that the model reproduces accurately dynamic characteristic between the mass air flow and the mass fuel gas flow and exhaust air fuel ratio in different operating conditions.