论文部分内容阅读
采用多层前向型神经网络，对燃煤循环流化床锅炉模型化进行了研究。以机理数学模型产生的数据样本对神经网络进行训练，结果表明，训练后的神经网络不仅可以精确地再现机理数学模型现有计算结果，而且可以比较精确地和机理模型一样对锅炉的性能进行预测。由于神经网络模型预测时间极短，并且可以不断随着新的数据样本进行自适应学习，从而为循环流化床锅炉的模型化及其实时应用（如实时训练仿真器、模型预测控制等）提供一个有效的新途径。
The multi-layer forward neural network was used to study the modeling of coal-fired circulating fluidized bed boiler. Training the neural network with the data samples generated by the mechanism mathematical model shows that the trained neural network can not only accurately reproduce the existing calculation results of the mechanism mathematical model but also predict the performance of the boiler more exactly as the mechanism model . Due to the extremely short predictive time of neural network model and its continuous learning with new data samples, it can be used to model the circulating fluidized bed boiler and its real-time applications such as real-time training simulator, model predictive control and so on An effective new way.