【作者】 张健；

【导师】 滕勤；

【作者基本信息】 合肥工业大学 ， 动力机械及工程， 2010， 硕士

【摘要】 电控技术是提高煤层气发动机燃烧效率、降低排放和提高动力性的重要手段。在煤层气发动机控制系统分析和设计过程中,空燃比动态建模和前馈MAP校正方法是需要研究的重要内容,也是电控单元控制软件优化设计和控制策略开发的基础。本文研究主要内容如下:(1)为了描述煤层气发动机空燃比的动态特性,建立了多变量空燃比块联模型。利用稳态工况实验数据拟合的多项式,补偿模型中的静态非线性增益。借助3个准则函数选择动态模型的阶次,基于发动机动态工况实验数据,分别采用预滤波的频率权值修正(SM)法和输出误差(OE)法辨识空燃比动态模型的参数。模型验证结果表明,基于SM算法的块联模型能够相对较好地捕获发动机空燃比的瞬态偏移。(2)为了描述煤层气发动机强烈的非线性和补偿动态工况下的延迟,基于减法聚类算法(SCA)和在线聚类算法(OCA),分别建立了用于空燃比反馈控制的ANFIS模型和自适应模糊模型。利用同向激励和反向激励下的动态数据,对两种模型预测空燃比的能力进行了检验和交叉验证。结果表明,尽管两种模型都能够补偿发动机各种延迟、描述排气空燃比的瞬态偏移,但基于OCA算法自适应模型,由于采用参数递推算法和适当选择的规则半径,通过在线修正和调整模型参数,可以更好地实时预测空燃比的变化,具有更高的精度。(3)为了优化模糊空间划分和进一步改善建模精度,考虑聚类中心的关联性,将协同系数引入G-K聚类算法中,结合系统性能指标和3种聚类数判定准则(SC、S、XB),构建了基于G-K协同聚类算法的发动机空燃比模糊模型。模型验证结果表明,与G-K聚类算法的模糊模型相比,基于G-K协同聚类算法的模糊模型具有更好的鲁棒性,更适合作为空燃比动态模型。(4)为了尽可能消除稳态控制偏差和优化发动机控制MAP,借助于辨识的发动机稳态模型,研究了基于PI型、PID型和自适应PID型迭代学习控制律的稳态空燃比自学习校正算法;为了补偿动态非线性和延迟引起的瞬态空燃比偏移,借助于基于协同聚类算法的空燃比动态模型,研究了基于PID型和自适应PID型迭代学习律的动态空燃比校正方法。仿真结果表明,自适应PID学习控制算法具有最快的收敛速度和最小的学习误差,更适合煤层气发动机位置控制参数的在线学习和调整;基于自适应PID学习控制律的方法通过快速的误差反馈学习和实时跟踪,具有更好的收敛性能和瞬态空燃比控制能力。更多还原

【Abstract】 The electronic control technique is an important way to heighten combustion efficiency, reduce exhaust gas and improve power performance for a coal-bed gas engine. During analysising and designing the control system, the study on dynamic air fuel ratio modeling method and feedforward MAP correcting method for a coal-bed gas engine are the key content, and it is also a basis for optimal design of control softwares and development of the control strategy in an electronic control unit.The main work as follows:(1) In order to describe the dynamic characteristic of air fuel ratio for the coal-bed gas engine, multivariable block-link model of the air fuel ratio was constructed. The polynomical fitted with the steady-state operation experiment data was used to compensate for static nonlinear gain of the model. By means of three kinds of criterione functions, the order of the dynamic model was chosen. Based on the dynamic operation experiment data, the parameters of the dynamic model were identified by the Steiglitz-McBride (SM) method and the output error method (OE). Model validation results show that a block-link model based on SM algorithm can capture transient excursion of air-fuel ratio more accurately.(2) In order to describe the strong non-linear characteristics and to compensate for the delay of the coal-bed gas engine during dynamic operation conditions, based on subtractive clustering algorithm (SCA) and online clustering algorithm (OCA), the ANFIS model and adaptive fuzzy model for air fuel ratio feedback control were constructed, respectively. Using dynamic operation experiment data with the same direction excitation and reverse direction excitation, the predicting capability of air fuel ratio was verified and cross-validated. Results show that although two models can compensate for the different engine delays and accurately describe transient excursion of exhaust air-fuel ratio, due to using recursive algorithms for parameters and appropriate choice for rule radius, and by means of online correction and adjustment, the adaptive model based on on-line clustering algorithm can better predict the real-time changes of air fuel ratio, and has higher accuracy.(3) In order to optimize fuzzy partition space and further improve the modeling precision, considering the relevance of cluster centers and introducing collaborative coefficients to G-K clustering algorithm, based on G-K collaborative clustering algorithm, the air fuel ratio fuzzy model was built with the combination of system performance index and three kinds of clustering evaluation criteria (SC、S、XB). The model validation results show that the fuzzy model based on G-K collaborative clustering algorithm has better robustness than that based on G-K clustering algorithm, and is more suitable as an air fuel ratio dynamic model.(4) In order to eliminate steady control diviation and optimize the engine control MAP as much as possible, by means of an identified steady-state model, the self-learning correction algorighms for steady-state air fuel ratio control were examined based on PI、PID and adaptive PID iterative lerning control law. To compensate for transient state excursion of air fuel ratio resulting from dynamic non-linearityand delays, with the help of an air fuel ratio dynamic model with G-K collaborative clustering algorithm, the dynamic correcting method for air fuel ratio were studied based on PID and adaptive PID iterative learning control law. Simulation results show that adaptive PID learning control algorithm has the fastest convergence rate and minimum learning error, and is much more suitable to online learning and adjusting position control parameters for the coal-bed gas engine. Through error feedback learning and real-time tracking, the method based on adaptive PID lerning congrol algorithm has better convergency and ability to control transient air fuel ratio.更多还原