基于RVM回归误差补偿的航空发动机分布式控制系统多步预测控制

针对具有随机有界双侧时延的航空发动机分布式控制系统,提出了一种基于多步预测和关联向量机(RVM)回归误差补偿的控制方案.首先建立航空发动机分布式控制系统(DCS)的神经网络非线性自回归滑动平均(NARMA)模型,利用当前的系统输出和控制量对N步之后的系统输出进行预测；其次用改进的RVM回归多步预测算法估计NARMA模型的的预测误差,并对预测结果进行误差补偿；最后利用补偿之后的预测值和设定值对控制参数进行滚动优化,设计系统的神经网络逆控制器实现系统的自适应控制.仿真结果证明该控制策略能够避免随机有界双侧时延对控制系统的影响,实现对设定值的稳定跟踪,且控制器具有较好的实时性和鲁棒性.低压转子转速阶跃响应的稳态绝对误差小于0.04％,响应时间小于0.3s.

Multi-step predictive control for aero-engine distributed control system based on RVM regression error compensation

A multi-step predictive control method based on relevance vector machine(RVM) regression error compensation was put forward to control aero-engine distributed control system(DCS) with stochastic and bounded time-delay.The neural network nonlinear auto regressive moving average(NARMA) model of the aero-engine DCS was established to predict the future output.An improved multi-step prognostic algorithm based on RVM regression was presented to estimate prognostic error series which were compensated for the NARMA model.Finally,a neural network inverse controller was designed,and rolling optimization of the parameters was conducted by prognostic output and the expectations.Simulation results illustrate the control strategy avoids the effect of stochastic and bounded time-delay on the control system.The controller has favorable dynamic tracking and robust performance.The absolute steady error of low-pressure rotor speed step response is less than 0.04%,and the response time is less than 0.3 s.