变速积分PID神经网络在CFB锅炉燃烧中的应用

Application of Shift Integral PID Neural Network Control Method in the Combustion Process of Circulating Fluidized Bed Boiler

对循环流化床锅炉燃烧系统中的给煤量-主汽压系统进行了分析。针对其大延迟、大惯性、非线性强且难以建立精确的数学模型的特点,以及传统PID控制中积分环节的不足,提出一种基于变速积分PID的神经网络控制。该方法将神经网络和PID控制相结合,既具有神经网络自学习、自适应及逼近任意函数的能力,又具有常规PID控制器结构简单的特点;同时,变速积分的引入,更加克服了针对于高度非线性、时滞性、对执行机构输出限制严格的系统,常规PID易产生积分饱和的问题。仿真结果表明,基于变速积分PID的神经网络控制器控制效果理想,能够对复杂系统实行有效控制。

The article analyses the feature of the system of the coal amount and the main steam pressure in the circulat-ing fluidized bed boiler. For its large delay, large inertia, strong nonlinear and difficult to establish accurate model features, as well as the deficiencies of the integral part in traditional PID, a control strategy based on shift integral PID neural network is proposed. This strategy combines the neural network and the PID control. It has the ability of self-learning, self-adapting, approximating any function as well as a simple structure. Also, the introduction of the shift in-tegral makes it better to overcome the problem of integral saturation in the systems with high nonlinearity, large time de-lay, and strict restriction on the output of actuator. The simulation result shows that control strategy based on shift inte-gral PID neural network is satisfactory, and it is able to implement effective control of complex systems.