摘要
考虑到热工对象的动态特性大多从阶跃响应获得,非线性的存在使此种线性动态模型只能在相应稳态点附近有良好的性能,文中采用了基于非线性稳态模型来实现动态模型自适应的策略,该策略用较为精确的非线性稳态模型得到当前输入下的稳态参数,然后由此修正线性动态模型中与稳态相关的参数,实现了动态模型的自适应,进而有效提高了大范围下的动态预测性能。通过对某电厂360MW"W型"火焰强制循环固态排渣煤粉炉的稳态和动态试验,建立了NOx的神经网络稳态模型和线性动态模型,用两个不同工况下的实际数据,验证了结合稳态模型的非线性自适应动态模型比线性动态模型具有更好的NOx排放预测性能。
As the dynamic characteristics of thermal objects are usually acquired by step responses, nonlinearity makes those models effective only near static work conditions, an approach is proposed to realize an adaptive dynamic model based on a nonlinear static model. First, steady state of current inputs is estimated by static model, then these real-time static estimate results are applied to update the parameters depending on steady state in linear dynamic models, after that an adaptive nonlinear dynamic model is achieved and this model can be used in a relative long-range dynamic prediction. Both static and dynamic experiments were carried out in a 360MV pulverized coal-fired boiler with W-type flame. Then, using ANN (Artificial Neural Network) a nonlinear static model was acquired based on 107 static data samples, and a series of linear dynamic models were obtained by 7 step response curves. Finally, these nonlinear adaptive model and linear model were used to predict NOx emissions at the same time in two dynamic processes at different loads, results show that the nonlinear adaptive model is much better than linear model to predict NOx emissions in dynamic process.
出处
《中国电机工程学报》
EI
CSCD
北大核心
2008年第29期93-98,共6页
Proceedings of the CSEE