基于神经网络优化模型的丝绵木瞬态液流模拟

Instantaneous sap flow velocity simulation of Euonymus bungeanus based on neural network optimization model

树木的液流规律是复杂的,难以用多元线性或经验模型表达,在理解林木树干液流规律的基础上,寻找一种简易可行的方法模拟林木树干液流对环境因子的响应过程,对定量分析森林生态水文过程及区域生态需水量尤为重要。本研究以宁夏河东沙区防护林树种丝绵木为对象,采用热扩散茎流计连续测定树干液流速率,分析环境因子对丝绵木树干液流的影响,并构建基于粒子群算法(PSO)和麻雀搜索算法(SSA)优化的神经网络模型对丝绵木液流速率进行预测。结果表明:影响丝棉木树干液流的主要因素为太阳辐射、饱和水汽压差、气温和相对湿度,重要度依次为32.5%、25.3%、22.0%和16.1%,其响应过程均呈现时滞回环关系。采用优化后的BP、Elman和ELM神经网络模型模拟瞬态液流,综合评价指标(GPI)分别提高1.5%、30.0%和5.3%。但与PSO-Elman和SSA-ELM优化模型相比,SSA-BP优化模型预测结果最佳,GPI分别提高1.0%和23.2%。基于麻雀搜索算法优化的BP神经网络模型可以作为预测丝绵木树干瞬态液流速率的首选模型。

The sap flow of trees is complex and difficult to express with multivariate linear or empirical models.A simple and feasible method on the basis of understanding sap flow variation to simulate its variation with environmental factors is of special importance for quantitatively analyzing forest ecohydrological processes and regional water demand.In this study,with one of the shelter forest species Euonymus bungeanus in the east sandy land of Yellow River in Ningxia as the research object,we continuously measured the trunk sap flow velocity by thermal diffusion sap flow meter,and analyzed the effects of environmental factors on stem sap flow.We used the particle swarm optimization(PSO)and sparrow search algorithm(SSA)optimized neural network model to predict sap flow velocity of E.bungeanus.Results showed that the main environmental factors influencing sap flow were solar radiation,vapor pressure deficit,air temperature,and relative humidity,with the influencing importance of 32.5%,25.3%,22.0%and 16.1%,respectively.The response process between sap flow and environmental factors presented a hysteresis loop relationship.The optimized BP,Elman and ELM neural network models improved the comprehensive evaluation index(GPI)by 1.5%,30.0%and 5.3%,respectively.Compared with the PSO-Elman and SSA-ELM optimization models,the SSA-BP optimization model had the best prediction results with an improvement of 1.0%and 23.2%in GPI,respectively.Therefore,the prediction results of the BP neural network model based on the sparrow search algorithm could be used as an optimal model for predicting instantaneous sap flow velocity of E.bungeanus.