论文部分内容阅读
针对传统的统计学方法难以精确刻画岩溶地下河日流量变化的非线性动态特性,引入有源自回归神经网络(NARX)技术,建立了基于NARX模型的岩溶地下河日流量预测模型,基于寨底地下河2013年1月15日~2014年6月30日的降雨量和流量数据,利用该模型对寨底地下河日流量进行了短期预测。结果表明,该模型预测效果较好,能够很好地预测岩溶地下河流量的变化趋势和极值等动态特性,另外该模型神经元个数越多,延迟阶数越大,神经网络对数据的学习能力和灵活性越强,但该模型不宜进行归一化处理。
According to the traditional statistical methods, it is difficult to accurately characterize the nonlinear dynamic characteristics of karst underground river daily flow, and the active recurrent neural network (NARX) is introduced to establish the daily flow forecasting model of karst underground river based on NARX model. Underground river from January 15, 2013 to June 30, 2014 rainfall and flow data, the use of the model to simulate the underground river daily flow at the end of a short-term forecast. The results show that this model has a good prediction effect and can well predict the dynamic characteristics such as trend and extreme value of karst underground river flow. In addition, the more neurons in this model, the larger the delay order, The stronger the learning ability and flexibility, but the model should not be normalized.