论文部分内容阅读
为定量研究BP人工神经网络不同输入层对径流模拟的影响,以滨江流域8个雨量监测站长系列逐日降水径流资料为基础,对比分析原始降水、算术平均降水、复合前期径流降水、流域面积加权降水和复合径流面积加权降水作为输入层时,BP人工神经网络月径流量模拟的结果差异。研究表明:采用流域面积加权降水模拟的径流量,具有最优相关系数和确定性系数,以原始降水作为输入层所得结果相对误差最小,由算术平均降水模拟出的结果分布最集中,网络模拟效果稳定。复合输入层的模拟精度相对较高,但输入层并非越复杂越好,基于面积加权降水得出的模拟径流量综合评价最高。
In order to quantitatively study the effects of different input layers of BP artificial neural network on runoff simulation, based on the daily rainfall runoff data of eight rainfall monitoring stations in Binjiang River Basin, the correlation analysis of primary precipitation, arithmetic average precipitation, pre-compound precipitation runoff, Precipitation and composite runoff area as the input layer of rainfall as the input, BP artificial neural network lunar flow simulation results. The results show that the runoff volume simulated by basin-area weighted precipitation has the best correlation coefficient and the certainty coefficient, and the relative error is the least with the initial precipitation as the input layer. The simulated result by arithmetic mean precipitation is the most concentrated. The network simulation effect stable. The simulation accuracy of the composite input layer is relatively high, but the input layer is not as complex as possible, and the simulation runoff yield based on the area weighted precipitation has the highest comprehensive evaluation.