| 摘要: |
| 【目的】精准模拟温室梨枣树液流量。【方法】基于粒子群算法（PSO）优化的极限学习机（ELM）模型，选取了西北旱区的温室梨枣树逐日气象资料和梨枣树生理指标作为输入参数，构建了16种不同参数组合的PSO-ELM模型对梨枣树各生育期的液流量进行模拟，并与实测液流值进行对比。【结果】PSO-ELM模型能通过较少的输入参数实现梨枣树液流量的高精度模拟：全生育期液流量模拟中M2模型（输入参数为叶面积指数、平均气温、实际水汽压、平均相对湿度、净辐射和风速）、M4模型（输入参数为叶面积指数、平均气温、实际水汽压、平均相对湿度、风速和土壤含水率）及M12模型（输入参数为叶面积指数、实际水汽压和平均相对湿度）的MAE、MBE、R2、MRE及RRMSE范围分别为1.467 6~1.598 6 mm/d、-0.000 9~0 mm/d、0.370 6~0.435 4、0.177 2~0.185 5及0.202 6~0.214 0，GPI排名分别1、2和5，其中M12的输入参数较少且模拟精度较高，其MAE、MBE、R2、MRE、RRMSE分别为1.598 6 mm/d、0、0.370 6、0.185 5、0.214 0；萌芽展叶期、开花坐果期、果实膨大期和果实成熟期液流量模拟结果分别以MⅠ-11模型（输入参数为净辐射、叶面积指数和实际水汽压）、MⅡ-15模型（输入参数为实际水汽压和平均气温）、MⅢ-11模型（输入参数为平均相对湿度、叶面积指数和土壤含水率）和MⅣ-12模型（输入参数为叶面积指数、净辐射和平均气温）模拟精度较高，GPI排名分别为8、2、4和5。【结论】PSO-ELM模型模拟温室梨枣树不同生育期液流量均具有较高的精度，可作为温室梨枣树液流量估算的新方法。 |
| 关键词: 液流量； 粒子群优化算法； 极限学习机； 温室； 梨枣树 |
| DOI：10.13522/j.cnki.ggps.2019012 |
| 分类号: |
| 基金项目: |
|
| Sap Flow of Pear-jujube Simulation in Greenhouse Based on PSO-ELM Model |
|
ZHANG Nian, CUI Ningbo, ZHAO Lu, XIAO Lu, ZHANG Fujuan, MA Zelong, YUE Jinhua
|
|
1. State Key Laboratory of Hydraulics and Mountain River Engineering & College of Water Resources and Hydro Power, Sichuan University, Chengdu 610065, China; 2. Institute of Water-saving Agriculture in Arid Areas of China, Northwest A&F University, Yangling 712100, China; 3.Provincial Key Laboratory of Water-saving Agriculture in Hill Area of Southern China, Chengdu 610066, China; 4. Sichuan Provincial Water Conservancy Research Institute, Chengdu 610072, China; 5. Beijing Dongfang Runze Ecological Technology Co., Ltd., Beijing 100191, China
|
| Abstract: |
| 【Objective】Accurately simulat the sap flow of pear-jujube in greenhouse.【Method】Based on the extreme learning machine (ELM) model of particle swarm algorithm (PSO) optimization, the daily meteorological data of pear-jujube in arid areas of Northwest China and the physiological index of pear-jujube tree were selected as input parameters, and 16 kinds of PSO-ELM models with different parameter combinations were constructed to simulate the sap flow of pear-jujube in each growth period and compared with the measured sap flow.【Result】The PSO-ELM model could realize the high precision simulation of pear-jujube sap flow with less input parameters: in total growth period, M2 model (the input parameters are leaf area index, average temperature, actual water vapor pressure, average relative humidity, net radiation and wind speed), M4 model (the input parameters are leaf area index, average temperature, actual water vapor pressure, average relative humidity, wind speed and soil moisture content) and M12 model (the input parameters are leaf area index, actual water vapor pressure and average relative humidity) had MAE, MBE, R2, MRE and RRMSE ranges of 1.467 6 to 1.598 6 mm/d, -0.000 9 to 0 mm/d, 0.370 6 to 0.435 4, 0.177 2 to 0.185 5 and 0.202 6 to 0.214 0, respectively, with GPI rankings of 1, 2 and 5 respectively, of which M12 had fewer input parameters but higher simulation accuracy, and its MAE, MBE, R2, MRE and RRMSE were 1.598 6 mm/d, 0, 0.370 6、0.185 5 and 0.214 0 respectively, and the results of sap flow simulation in the reproductive period showed that the MⅠ-11 model (the input parameters are net radiation、leaf area index and actual water vapor pressure) were used in the germination period, flowering fruit sitting period, fruit expansion period and fruit ripening period respectively, the simulation accuracy of MⅡ-15 model (the input parameters are actual water vapor pressure and average temperature), MⅢ-11 model (the input parameters are average relative humidity、leaf area index and soil moisture content) and MⅣ-12 model (the input parameters are leaf area index, net radiation and average temperature) were high, whose GPI rankings were 8, 2, 4 and 5, respectively. 【Conclusion】The simulation of sap flow in different growth periods of pear-jujube in PSO-ELM model had high accuracy, which could be a new method for estimating the sap flow of pear-jujube in greenhouse. |
| Key words: sap flow; particle swarm optimization algorithm; extreme learning machine; greenhouse; pear-jujube |
