基于机器学习方法的“以电折水”系数研究  

Research on Coefficient of“Converting Electricity into Water”based on Machine Learning Method

作　　者：冯佳伟 姜宁[1] 刘岩 董万里 杨莹 Feng Jiawei;Jiang Ning;Liu Yan;Dong Wanli;Yang Ying(College of Water Resources and Electric Power,Heilongjiang University,Harbin 150080,China;Heilongjiang Provincial Water Conservancy Science Research Institute,Harbin 150080,China)

机构地区：[1]黑龙江大学水利电力学院,哈尔滨150080 [2]黑龙江省水利科学研究院,哈尔滨150080

出　　处：《黑龙江水利科技》2025年第3期139-144,共6页Heilongjiang Hydraulic Science and Technology

基　　金：国家重点研发计划(2022YFD1500402)。

摘　　要：目前“以电折水”作为农业地下水开采量的间接计量方法,估算地区农灌机井“以电折水”系数对于地区地下水开采量准确计量至关重要。当前多数研究采用在灌溉机井出水口安装计量设施,直接测量机井单位时间内的抽水量与耗电量来计算“以电折水”系数的具体数值。相比之下,利用数学模型并结合相关影响因素对“以电折水”系数进行预测的研究较为稀缺。文章针对“以电折水”系数进行预测及分析研究,采用平均相对误差(MRE)、均方根误差(RMSE)以及决定系数(R^(2))对不同模型的预测准确性展开对比。研究表明,采用平均值法预测的“以电折水”系数误差平均为30.33%,表明该方法下的预测结果与实际数据之间的拟合度较低,精度欠佳;而使用径向基函数神经网络模型时,预测误差则降低至11.23%;支持向量机回归模型展现了良好的预测性能,其预测误差仅为9.29%,显示出最佳的数据拟合度与最高的预测精度。At present,“estimating water consumption from electricity consumption”is an indirect method for measuring agricultural groundwater extraction,estimating the“estimating water consumption from electricity consumption”coefficient of agricultural irrigation wells in a region is crucial for accurate measurement of the groundwater extraction in that region.Most current studies calculate the specific value of the“estimating water consumption from electricity consumption”coefficient by installing measurement facilities at the outlet of irrigation wells to directly measure the pumping volume and power consumption of the wells per unit time.In contrast,research on predicting the“estimating water consumption from electricity consumption”coefficient using mathematical models and considering relevant influencing factors is relatively scarce.This paper conducts research on the prediction and analysis of the“estimating water consumption from electricity consumption”coefficient,and compares the prediction accuracy of different models using the Mean Relative Error(MRE),Root Mean Square Error(RMSE),and Coefficient of Determination(R^(2)).The research shows that the average error of the“estimating water consumption from electricity consumption”coefficient predicted by the average value method is 30.33%,indicating a low degree of fit between the prediction results and the actual data under this method,and poor accuracy.When using the Radial Basis Function Neural Network model,the prediction error is reduced to 11.23%.The Support Vector Machine Regression model demonstrates excellent prediction performance,with a prediction error of only 9.29%,showing the optimal data fit and the highest prediction accuracy.

关 键 词：“以电折水”系数 径向基函数神经网络模型 支持向量机回归模型 农业地下水开采计量 

分 类 号：S274.4[农业科学—农业水土工程]