Isolated Area Load Forecasting using Linear Regression Analysis: Practical Approach

This paper presents an analysis to forecast the loads of an isolated area where the history of load is not available or the history may not represent the realistic demand of electricity. The analysis is done through linear regression and based on the identification of factors on which electrical load growth depends. To determine the identification factors, areas are selected whose histories of load growth rate known and the load growth deciding factors are similar to those of the isolated area. The proposed analysis is applied to an isolated area of Bangladesh, called Swandip where a past history of electrical load demand is not available and also there is no possibility of connecting the area with the main land grid system.

Statistical analysis of nitrogen use efficiency in Northeast China using multiple linear regression and Random Forest

Understanding the spatial-temporal dynamics of crop nitrogen(N)use efficiency(NUE)and the relationship with explanatory environmental variables can support land-use management and policymaking.Nevertheless,the application of statistical models for evaluating the explanatory variables of space-time variation in crop NUE is still under-researched.In this study,stepwise multiple linear regression(SMLR)and Random Forest(RF)were used to evaluate the spatial and temporal variation of NUE indicators(i.e.,partial factor productivity of N(PFPN);partial nutrient balance of N(PNBN))at county scale in Northeast China(Heilongjiang,Liaoning and Jilin provinces)from 1990 to 2015.Explanatory variables included agricultural management practices,topography,climate,economy,soil and crop types.Results revealed that the PFPN was higher in the northern parts and lower in the center of the Northeast China and PNBN increased from southern to northern parts during the 1990–2015 period.The NUE indicators decreased with time in most counties during the study period.The model efficiency coefficients of the SMLR and RF models were 0.44 and 0.84 for PFPN,and 0.67 and 0.89 for PNBN,respectively.The RF model had higher relative importance of soil and climatic covariates and lower relative importance of crop covariates compared to the SMLR model.The planting area index of vegetables and beans,soil clay content,saturated water content,enhanced vegetation index in November&December,soil bulk density,and annual minimum temperature were the main explanatory variables for both NUE indicators.This is the first study to show the quantitative relative importance of explanatory variables for NUE at a county level in Northeast China using RF and SMLR.This novel study gives reference measurements to improve crop NUE which is one of the most effective means of managing N for sustainable development,ensuring food security,alleviating environmental degradation and increasing farmer’s profitability.

Prediction and driving factors of forest fire occurrence in Jilin Province,China

Forest fires are natural disasters that can occur suddenly and can be very damaging,burning thousands of square kilometers.Prevention is better than suppression and prediction models of forest fire occurrence have developed from the logistic regression model,the geographical weighted logistic regression model,the Lasso regression model,the random forest model,and the support vector machine model based on historical forest fire data from 2000 to 2019 in Jilin Province.The models,along with a distribution map are presented in this paper to provide a theoretical basis for forest fire management in this area.Existing studies show that the prediction accuracies of the two machine learning models are higher than those of the three generalized linear regression models.The accuracies of the random forest model,the support vector machine model,geographical weighted logistic regression model,the Lasso regression model,and logistic model were 88.7%,87.7%,86.0%,85.0%and 84.6%,respectively.Weather is the main factor affecting forest fires,while the impacts of topography factors,human and social-economic factors on fire occurrence were similar.

Performance of the geometric approach to fault detection and isolation in SISO,MISO,SIMO and MIMO systems

In this paper,a geometric approach to fault detection and isolation (FDI) is applied to a Multiple-Input Multi-ple-Output (MIMO) model of a frame and the FDI results are compared to the ones obtained in the Single-Input Single-Output (SISO),Multiple-Input Single-Output (MISO),and Single-Input Multiple-Output (SIMO) cases. A proper distance function based on parameters obtained from parametric system identification method is used in the geometric approach. ARX (Auto Regressive with eXogenous input) and VARX (Vector ARX) models with 12 parameters are used in all of the above-mentioned models. The obtained results reveal that by increasing the number of inputs,the classification errors reduce,even in the case of applying only one of the inputs in the computations. Furthermore,increasing the number of measured outputs in the FDI scheme results in decreasing classification errors. Also,it is shown that by using probabilistic space in the distance function,fault diagnosis scheme has better performance in comparison with the deterministic one.

基于图像处理的水培生菜冠层图像叶面积估测研究

为实现精准、高效、无损地获取植物工厂环境下水培生菜相关长势参数叶面积(Leaf area,LA),基于数字图像处理和机器学习回归方法建立单株水培生菜冠层图像LA估测模型。首先,通过智能手机获取2个生菜品种不同生长期的冠层可见光图像,利用Photoshop图像处理软件将原始图像统一剪裁为900像素×900像素大小,采用中值滤波(MedianBlur)法对剪裁后的图像进行去噪运算,将RGB图像转化为HSV颜色空间,再采用mask掩膜法分割彩色图像;然后,利用图像法获取单株生菜LA实测值,构建以LA实测值为因变量,以生菜冠层投影面积(Projected leaf area,PLA)为自变量的线性回归(Linear regression,LR)模型和以全局图像特征(颜色、形状、纹理等)为自变量的支持向量回归(Support vector regression,SVR)、多元线性回归(Multiple linear regression,MLR)和随机森林(Random forest,RF)等LA估测模型进行对比分析;最后,采用决定系数(Coefficient of determination,R^(2))和均方根误差(Root mean square error,RMSE)评估模型的准确性。结果表明:RF模型估测效果最好,对于生菜品种‘绿萝’单株LA估测结果的R^(2)为0.9714、RMSE为8.89 cm2,对于品种‘碧霄’估测结果的R^(2)为0.9201、RMSE为23.34 cm2。本研究验证了RF回归模型能够较准确地估测生菜单株叶面积,可为植物工厂水培生菜LA无损估测提供新的解决方案和研究基础。

基于无人机多光谱影像的云南松林蓄积量估测模型

【目的】无人机多光谱遥感影像较可见光影像具有更丰富的光谱信息,在森林蓄积量估测中具有较大潜力。以无人机载多光谱遥感影像为主要数据源,探索森林蓄积量的遥感估测模型,以克服传统地面调查工作量大、耗时长、成本高等弊端。【方法】以滇中地区典型天然云南松Pinusyunnanensis纯林为研究对象,利用无人机多光谱影像提取单波段反射率、各类植被指数、纹理特征等,计算各特征变量的标准地均值;筛选与云南松林蓄积量相关性显著的特征变量,采用多元线性、随机森林、支持向量机建立云南松林蓄积量估测模型,以决定系数(R^(2))、平均绝对误差(E_(MA))、均方根误差(E_(RMS))、平均相对误差(EMR)评价模型精度。【结果】①3种模型中,随机森林的精度最高(R^(2)=0.89,E_(MA)=4.69 m^(3)·hm^(-2),E_(RMS)=5.45 m^(3)·hm^(-2),EMR=14.5%),其次为支持向量机(R^(2)=0.74,E_(MA)=5.27 m^(3)·hm^(-2),E_(RMS)=8.31 m^(3)·hm^(-2),EMR=13.1%),最低为多元线性回归模型(R^(2)=0.35,E_(MA)=10.12 m^(3)·hm^(-2),E_(RMS)=12.85 m^(3)·hm^(-2),EMR=28.1%);3种模型在测试集上的估测精度均有所降低,随机森林的模型表现最好,支持向量机次之,多元线性最差。②3种模型在云南松林蓄积量估测中均存在一定的低值高估和高值低估现象。③基于无人机多光谱影像估测云南松林蓄积量,纹理特征仍是不可忽视的重要因子。【结论】基于无人机多光谱影像,在不进行单木分割的情景下,提取标准地的单波段反射率、植被指数、纹理特征均值,筛选适用于蓄积量估算的变量构建估测模型。通过对3种模型进行精度评价,随机森林为云南松林蓄积量估测的最佳模型。

基于特征光谱参数的叶片和冠层尺度茶多酚含量估算

茶多酚具有很强的生理活性和抗氧化性,是茶品质的重要属性之一。相比传统茶多酚含量的测定方法,遥感技术监测茶多酚含量具有高效、精确及实时的优势,但如何利用遥感数据监测不同时期的茶多酚含量研究较少。该研究以广东省英德市的5个茶园的茶叶为研究对象,对春茶、夏茶和秋茶的叶片与冠层两个尺度的茶多酚含量及对应高光谱数据进行测定,利用标准正态变量变换(SNV)对叶片和冠层的高光谱反射率数据进行预处理;然后,分别采用连续投影算法(SPA)和竞争性自适应重加权采样算法(CARS)筛选不同生长季节叶片和冠层两个尺度茶多酚的敏感波段;最后,通过偏最小二乘法(PLS)、随机森林(RF)和多元线性回归(MLR)分别构建不同时期的茶多酚含量模型并进行验证。结果表明:(1)茶多酚的含量随着季节推移显著增加,春茶茶多酚含量(15.37%)最低,夏茶茶多酚含量次之(18.29%),秋茶茶多酚含量(秋茶20.77%)最高;(2)不同敏感波段筛选的茶多酚含量的光谱特征波段主要为2100~2200 nm附近、1300~1400 nm附近、红波-红边波段及绿波段;(3)基于春茶、夏茶和秋茶冠层光谱特征构建的茶多酚模型中CARS-PLS、SPA-MLR和CARS-PLS模型精度最高,建模集R2分别为0.56、0.45和0.52,RMSE分别为1.15、1.68和1.77;验证集R2分别为0.43、0.40和0.41,RMSE分别为1.60、1.91和1.91;基于春茶、夏茶和秋茶冠层叶片光谱特征构建的茶多酚模型中SPA-PLS、CARS-PLS和SPA-MLR模型精度最高,建模集R2分别为0.50、0.42和0.42,RMSE分别为1.25、1.70和1.66;验证集R2分别为0.43、0.36和0.38,RMSE分别为1.44、1.96和2.49。研究结果表明,基于遥感数据进行不同季节的叶片和冠层两个尺度的茶多酚含量估算是可行的,在大面积实时监测茶品质特征方面具有较大的潜力。

基于激光诱导击穿光谱(LIBS)的煤灰熔点快速检测

由于劣质煤的存在,燃煤时锅炉内极易出现结渣问题,不仅降低了煤的利用率、电厂的经济效益,且使火电厂存在安全隐患,如管壁超温爆管等,不符合国家节能减排的发展战略要求。因此炉内结渣是影响火电机组和气化工艺可靠运行的关键因素之一,准确预测灰熔点可以提前调整炉膛出口温度以避免结渣。采用激光诱导击穿光谱仪(LIBS)对33个火电厂常用煤燃烧后得到的煤灰样品采集其中所有元素的光谱,分别建立煤灰中的所有元素的谱线强度与煤灰熔点的随机森林模型、支持向量机回归模型和线性回归模型,直接预测煤灰熔点温度。采用基于马氏距离(MD)的异常数据剔除算法和基于稀疏矩阵的基线估计与降噪算法(BEADS),对粉煤灰样的全光谱数据进行了预处理。将全部数据划分为70%训练集和30%测试集,并以平均相对误差(MRE)作为判断模型拟合程度的标准,其中随机森林模型对粉煤灰熔点的预测平均相对误差(MRE)为54.74%,支持向量机回归模型的预测平均相对误差为60.08%,而线性回归模型的预测平均相对误差达到了9.78%。研究结果表明,线性回归模型对煤灰熔点的预测结果更准确。将便携式LIBS光谱仪与机器学习算法相结合,用于煤灰熔点的快速检测,可有效避免锅炉结渣问题,实现火电厂安全性和经济效益的同步提高,有广阔的发展前景。

基于RF和MLR的土壤重金属影响因素分析及生物有效性预测

为探究影响土壤中重金属累积和生物有效性的因素,以桂北地区某铅锌矿流域为研究对象,综合运用单因子指数法、风险评价编码法(RAC)、多元线性回归模型(MLR)和随机森林模型(RF)进行土壤重金属(Pb、Zn、Cu和Cr)累积影响因素分析及生物有效性预测。结果表明:研究区Cr含量无超标且空间分布相对均匀(变异系数为0.51);Cu、Pb和Zn的含量均值(分别为52.58、280.31 mg·kg^(-1)和654.71 mg·kg^(-1))均大于广西西江流域土壤重金属背景值,在思的河山前和地下河入口处全量和生物有效性均较大,对土壤生态环境具有一定风险;对于重金属全量分布和生物有效态的影响因素,阳离子交换量(CEC)、黏粒(Clay)、土壤有机质(SOM)和铁铝氧化物对Cr影响较大,SOM、Clay、pH和铁铝氧化物对Cu影响较大,pH、电导率(EC)和Clay对Pb影响较大,CEC、pH、土壤质地和铁铝氧化物对Zn影响较大;生物有效性预测结果显示RF和MLR均可较好地预测土壤重金属的全量与次生相,其中RF预测的R2区间为0.44~0.93,MLR预测的R2区间为0.30~0.72,RF预测结果表现更为准确。

基于逐步多元线性回归和随机森林模型预测黄河流域极端气温事件

全球变暖背景下,极端气候事件频发,且对黄河流域等地区的经济发展及人民生活造成严重危害。基于1961—2020年黄河流域80个站点的日气温数据提取了6个逐月极端气温指数(ETI)。利用多重共线性分析去除有相依性的环流指数,并考虑滞后性进行Pearson相关分析,筛选出各ETI的关键环流指数及最佳滞后时间;之后基于最佳滞后时间下的关键环流指数建立逐步多元线性回归(SMLR)和随机森林(RF)模型。对模型进行精度评价,探究环流指数在单站点及整个流域的重要性,并预测了2022年11月的6个ETI值。结果表明:黄河流域ETI中最高气温(TXx)、暖昼天数(TX90p)、酷热天数(TD30)和最低气温(TNn)呈波动上升趋势,而霜冻天数(FD0)和冷夜天数(TN10p)呈下降趋势;极端高温事件的强度和发生频率的空间分布特征与极端低温事件基本相反。以靖远站TXx为例,各关键环流指数对TXx具有不同程度的影响(0.10<r_(max)<0.89),r_(max)对应的最佳滞后时间主要为5、6、11、12个月。SMLR和RF模型对黄河流域各ETI的预测能力都较好,验证期的决定系数(R 2)范围分别为0.53~0.95和0.64~0.95;除对TXx的模拟效果稍弱外,其他5个ETI的RF模型模拟效果均优于SMLR模型。太平洋区极涡强度指数(PPVI)是影响黄河流域TXx、TNn、TX90p和FD0的最重要环流因子,北非—北大西洋—北美副高脊线位置指数(NANRP)对TN10p和TD30的影响最大。预测的2022年11月ETI的空间分布特征与多年平均情况基本相似。研究结果为黄河流域极端气温事件预报提供了参考。

数字经济时代林产品贸易与碳汇市场的相互作用——基于回归模型分析

数字经济时代,林产品贸易与碳汇市场的相互作用日趋显著。本文深入探讨了林产品贸易与碳汇市场的相互作用,并利用多元线性回归模型分析其对碳排放量的影响。研究发现,林业固定资产投资、人口总数、国内生产总值、林产品总产量和进口量对碳排放有显著影响。同时,本文探索了数字化技术如物联网、人工智能和区块链在提高林产品贸易效率和促进环境友好型发展方面的作用,这些技术的应用不仅优化了贸易流程,还提高了资源利用效率和环保标准。研究结果为林产品贸易的可持续发展和政策制定提供了重要依据,表现为数字经济时代下,运用先进技术和创新管理策略对减少环境影响的必要性。

Economic contribution of participatory agroforestry program to poverty alleviation: a case from Sal forests, Bangladesh

In the Forest Department of Bangladesh, a Participatory Agroforestry Program （PAP） was initiated at a denuded Sal forests area to protect the forest resources and to alleviate poverty amongst the local poor population. We explored whether the PAP reduced poverty and what factors might be responsible for poverty alleviation. We used three poverty measurement methods： the Head Count Index, the Poverty Gap Index and the Foster-Greer-Thorbecke index to determine the extent poverty reduction. We used a linear regression model to determine the possible differences among factors in poverty reduction. Data were collected through semi structured questionnaires and face to face interviews within the study area. PAP proved effective at poverty alleviation, considerably improving the local situation. The linear regression model showed that PAP output explained the income differences in poverty reduction. Participants identified bureaucracy and illegal money demands by forest department officials, an uncontrolled market system, and underdeveloped road infrastructure as the main obstacles to reduction of poverty. Overall, PAP is quite successful in alleviating poverty. So this program might be of interest at other degraded forest areas as a tool to alleviate poverty.

Spatiotemporal variation in forest fire danger from 1996 to 2010 in Jilin Province,China

We evaluated the spatial and temporal patterns of forest fires in two fire seasons (March to June and September to November) from 1996 to 2010 in Jilin Province, China, using the Canadian Forest Fire Weather Index System. Fire data were obtained from the Provincial Fire Agency, and historical climate records of daily weather observations were collected from 36 weather stations in Jilin and its neighboring provinces. A linear regression model was used to analyze linear trends between climate and fire weather indices with time treated as an independent variable. Correlation analysis was used to detect correlations between fire frequency, areas burned, and fire weather indices. A thin-plate smooth spline model was used to interpolate the point data of 36 weather stations to generate a surface covering the whole province. Our analyses indicated fire frequency and areas burned were significantly correlated with fire weather indices. Overall, the Canadian Forest Fire Weather Index System appeared to be work well for determining the fire danger rating in Jilin Province. Also, our analyses indicated that in the forthcoming decades, the overall fire danger in March and April should decrease across the province, but the chance of a large fire in these months would increase. The fire danger in the fall fire season would increase in the future, and the chance of large fire would also increase. Historically, because most fires have occurred in the spring in Jilin Province, such a shift in the future fire danger between the two fire seasons would be beneficial for the province's fire management.

Machine learning-based estimates of aboveground biomass of subalpine forests using Landsat 8 OLI and Sentinel-2B images in the Jiuzhaigou National Nature Reserve,Eastern Tibet Plateau

Accurate estimates of forest aboveground biomass(AGB)are critical for supporting strategies of ecosystem conservation and climate change mitigation.The Jiuzhaigou National Nature Reserve,located in Eastern Tibet Plateau,has rich forest resources on steep slopes and is very sensitive to climate change but plays an important role in the regulation of regional carbon cycles.However,an estimation of AGB of subalpine forests in the Nature Reserve has not been carried out and whether a global biomass model is available has not been determined.To provide this information,Landsat 8 OLI and Sentinel-2B data were combined to estimate subalpine forest AGB using linear regression,and two machine learning approaches–random forest and extreme gradient boosting,with 54 inventory plots.Regardless of forest type,Observed AGB of the Reserve varied from 61.7 to 475.1 Mg hawith an average of 180.6 Mg ha.Results indicate that integrating the Landsat 8 OLI and Sentinel-2B imagery significantly improved model efficiency regardless of modelling approaches.The results highlight a potential way to improve the prediction of forest AGB in mountainous regions.Modelled AGB indicated a strong spatial variability.However,the modelled biomass varied greatly with global biomass products,indicating that global biomass products should be evaluated in regional AGB estimates and more field observations are required,particularly for areas with complex terrain to improve model accuracy.

Forest Carbon Storage and Tree Carbon Pool Dynamics under Natural Forest Protection Program in Northeastern China

The Natural Forest Protection(NFP) program is one of the Six Key Forestry Projects which were adopted by the Chinese Government since the 1980s to address important natural issues in China. It advanced to protecting and restoring the structures and functions of the natural forests through sustainable forest management. However, the role of forest carbon storage and tree carbon pool dynamics since the adoption of the NFP remains unknown. To address this knowledge gap, this study calculated forest carbon storage(tree, understory, forest floor and soil) in the forest region of northeastern(NE) China based on National Forest Inventory databases and field investigated databases. For tree biomass, this study utilized an improved method for biomass estimation that converts timber volume to total forest biomass; while for understory, forest floor and soil carbon storage, this study utilized forest type-specific mean carbon densities multiplied by their areas in the region. Results showed that the tree carbon pool under the NFP in NE China functioned as a carbon sink from 1998 to 2008, with an increase of 6.3 Tg C/yr, which was mainly sequestrated by natural forests(5.1 Tg C/yr). At the same time, plantations also acted as a carbon sink, reflecting an increase of 1.2 Tg C/yr. In 2008, total carbon storage in forests covered by the NFP in NE China was 4603.8 Tg C, of which 4393.3 Tg C was stored in natural forests and 210.5 Tg C in planted forests. Soil was the largest carbon storage component, contributing 69.5%–77.8% of total carbon storage; followed by tree and forest floor, accounting for 16.3%–23.0% and 5.0%–6.5% of total carbon storage, respectively. Understory carbon pool ranged from 1.9 to 42.7 Tg C, accounting for only 0.9% of total carbon storage.

The effect of climate factors on the size of forest wildfires(case study:Prague-East district,Czech Republic)

This paper presents a new approach to identifying the climate variables that influence the size of the area burned by forest wildfires.Multiple linear regression was used in combination with nonlinear variable transformations to determine relevant nonlinear forest wildfire size functions.Data from the Prague-East District of the Czech Republic was used for model derivation.Individual burned forest area was hypothesized as a function of water vapor pressure,air temperature and wind speed.Wind speed was added to enhance predictions of the size of forest wildfires,and further improvements to the utility of prediction methods were added to the regression equation.The results show that if the air temperature increases,it may contain less water and the fuel will become drier.The size of the burned area then increases.If the relative humidity in the air increases and the wind speed decreases,the size of the burned area is reduced.Our model suggests that changes in the climate factors caused by ongoing climate change could cause significant changes in the size of wildfire in forests.

GPS- vs. DEM-Derived Elevation Estimates from a Hardwood Dominated Forest Watershed

Topographic attributes are often used as explanatory variables when providing spatial estimates of various environmental attribute response variables. Elevation of sampling locations can be derived from global positioning systems (GPS) or digital elevation models (DEM). Given the potential for differences in elevation among these two data sources, especially in response to forest canopy cover, our objective was to compare GPS and DEM-derived elevation values during the dormant season. A non-parametric Wilcoxon test indicated GPS elevation was higher than DEM elevation with a mean difference of 6 m. Linear regression analysis indicated that GPS and DEM elevation were well correlated (R2 = 0.71, r = 0.84, p 【0.0001). Although elevation among the two data sources differed, the strong linear relationship allows for correction of elevation values in a predictable manner.

基于多距离度量kNN模型的森林蓄积量反演

【目的】森林蓄积量是衡量森林质量和生长状况的重要指标。利用遥感技术进行森林蓄积量反演相比传统的森林调查能显著提高森林资源调查效率,对快速获取区域范围森林生长状况,进行高效的资源利用和森林经营管理具有重要意义。【方法】以陕西韩城市为研究区,利用森林资源二类调查数据库提取森林蓄积量实测数据,结合Sentinel-2遥感影像进行森林蓄积量反演。通过线性逐步回归法和重要性评价法分别进行变量筛选,构建多元线性回归模型、支持向量机模型、随机森林模型和基于欧式距离、曼哈顿距离和马氏距离构建的kNN模型进行森林蓄积量估测,通过精度评价比较最终选择估测精度最高的模型进行研究区森林蓄积量反演。【结果】1)马氏距离是最适合构建kNN模型的距离度量。基于马氏距离构建的kNN模型在所有模型中实现了最高的估测精度,决定系数R2为0.66,均方根误差RMSE为10.02 m^(3)/hm^(2),均方根误差相比随机森林模型、支持向量机模型和多元线性回归分别下降了3.9%、7.8%和29.9%;2)非参数模型在森林蓄积量估测中的精度显著优于参数模型。基于马氏距离构建的kNN模型、随机森林模型、支持向量机模型均方根误差相比多元线性回归分别降低了29.9%、27.0%和23.9%;3)研究区西北部森林生长情况较好,蓄积量值较大,东部和南部地区主要是水域和建筑用地,森林分布较少,森林蓄积量值较低。【结论】利用kNN模型结合Sentinel-2遥感影像能实现森林蓄积量反演和制图,为森林资源遥感估测研究提供参考。

应用机器学习模型与线性模型预测森林蓄积生长量的精度

为了探究不同模型对森林蓄积生长量的预测精度和影响因素,以吉林省汪清金沟岭林场为研究区,利用林场198个固定样地(20 m×20 m)中的乔木数据,建立多元线性回归(MLR)、随机森林(RF)、支持向量机(SVM)、最近邻法(KNN)模型,分析树木蓄积生长量与林分因子、地形因子和气候因子的关系,并比较不同模型的预测精度。结果表明:森林蓄积生长量与林分每公顷断面积(B_(A))、大于对象木断面积(B_(L))、海拔(A_(L))、林分密度(N)、年均降水量(P)显著相关,与平均胸径(D)、坡向(A_(s))、坡度(S_(L))、年均气温(T)并不相关;不同模型的预测结果存在差异,随机森林(RF)方法所有测试指标最佳;随机森林方法得出的平均蓄积生长量预测值为51.45 m^(3)·hm^(-2),该模型均方根误差(R_(MSE))最低(4.62),决定系数(R2)最高(0.91)。

基于机器学习算法的蒸发量模型评估

[目的]探讨合理的气候因子个数,建立蒸发量模型,提出基于特征选择算法筛选最优特征集。[方法]以陕西榆林、泾河和汉中3个气象站16年(2005-01至2021-03)的逐小时观测资料为研究对象,利用特征选择函数和遍历循环方法对模型参数、特征变量个数进行优化。基于最佳参数结合随机森林模型和多元线性回归模型两种机器学习算法建立榆林、汉中和泾河地区蒸发量模型,采用平均绝对误差、均方根误差和平方相关系数三项指标评估模型的预测精度。[结果]特征变量和随机森林模型中的决策树个数分别是8,61时,模型预测效果最佳。采用优化的随机森林模型、多元线性回归模型评估3个地区的平均绝对误差均为0,均方根误差除泾河地区相等外,榆林、汉中地区的均方根误差均小于优化的多元线性回归模型。优化的随机森林模型预测榆林、泾河和汉中地区蒸发量拟合效果分别为0.85,0.90,0.86,优化的多元线性回归模型的拟合效果分别为0.77,0.83,0.79。[结论]整体而言,优化的两种模型都具有良好的预测效果且随机森林模型的预测效果优于多元线性回归模型。