本文介绍了一种基于神经网络的热舒适控制器研究,建立了热舒适区域模型,最后对一VAV系统进行了模拟。
Abstract—This paper describes the design of a thermal comfort controller for indoor thermal environment regulation. In this controller, Predicted Mean Vote (PMV) is adopted as the
control objective and six variables are taken into consideration. Meanwhile, a kind of direct neural network (NN) control is designed, and a thermal space model for Variable-Air-Volume (VAV) application is developed. Based on the computer simulation, it is seen that this thermal comfort controller can maintain the indoor comfort level within the desired range
under both heating / cooling modes. Furthermore, by combining the energy saving strategy with the VAV application, it also shows the potential for energy saving in future.
应用误差反向传播的人工神经网络技术对大跨度双层扭网壳进行设计,双层扭网壳的跨度在50~ 80 m 之间变化,训练了估算双层扭网壳的最大挠度、重量和杆单元横截面面积的神经网络。为减少数据的非线性和提高训练速度,形成了特殊的数据排序方法。这种方法提供了必要的稳定性。本文研究表明应用神经网络技术对双层网壳结构进行设计是可行的。
本文在总结大量洪水预报实践经验的基础上,提出了一种峰值识别理论及相应的改进BP算法(Error Back Propagation with Peak Recognizer,简称BPPR).该理论及算法在修改网络权重时,偏重大值误差,即大值误差对权重的修改起主要作用.这种BPPR算法使人工神经网络洪水预报模型对洪峰的预报精度显著提高,从而保证了洪峰预报的可靠性.