基于GA-SA-SVR模型的输电边坡危险性预测  

Hazard Prediction of Transmission Slope Based on GA-SA-SVR Model

作　　者：段国勇 韩亮 王彦海[1,2] 吕军旗 郑武略 DUAN Guo-yong;HAN Liang;WANG Yan-hai;LYU Jun-qi;ZHENG Wu-lue(Hubei Transmission Line Engineering Technology Research Center,Three Gorges University,Yichang 443002,China;School of Electrical and New Energy,Three Gorges University,Yichang 443002,China;China Souther Power Grid Co.,Ltd.,Guangzhou Bureau,EHV Transmission Company,Guangzhou 510600,China)

机构地区：[1]湖北省输电线路工程技术研究中心(三峡大学),湖北宜昌443002 [2]三峡大学电气与新能源学院,湖北宜昌443002 [3]中国南方电网有限责任公司超高压输电公司广州局,广东广州510600

出　　处：《计算机技术与发展》2024年第12期207-212,共6页Computer Technology and Development

基　　金：国家自然科学基金联合基金项目(U2034203,U22A20600);南方电网广州局输电线路塔基区域边坡灾害危险性评价与分级研究技术服务科研项目(SDHZ2022341)。

摘　　要：输电边坡的危险性预测对于国内输电线路安全运营意义重大。该文依托某输电公司输电线路区域的边坡隐患排查及状态评估数据,对数据库进行筛选,以杆塔边缘距离、边坡高度、坡度、周边土地情况、岩土性质以及植被情况这六项作为输入特征值,危险系数作为输出标签建立支持向量回归(Support Vector Regression, SVR)预测模型,并采用遗传(Genetic Algorithm, GA)和模拟退火(Simulated Annealing algorithm, SA)的单独优化算法和组合优化算法分别对SVR模型进行优化,并设置鱼鹰、猎食者等优化算法作为对照组。结果表明:组合算法的优化效果要优于单一算法的优化效果,遗传-模拟退火组合算法(GA-SA)的优化效果在准确率和拟合程度上更有优势,测试集R2为0.937 5,MSE值为0.001 2,适应度函数f(x)值为0.072 4。该模型预测性能较好,相较原方法更加客观智能。Hazard evaluation of transmission slopes is of great significance for the safe operation of transmission lines in China.Relying on the data of transmission line slope hazard investigation and condition assessment of a transmission company,the database was screened,and the six items of distance from the edge of the tower,height of the slope,slope,surrounding land,geotechnical properties,and vegetation were used as the input eigenvalues,and the hazard coefficients were used as the output labels to establish a prediction model using Support Vector Regression(SVR).The Genetic Algorithm(GA)and Simulated Annealing(SA)of the individual optimization algorithm and the combination of optimization algorithms are used to optimize the SVR model,respectively,and set up OOA,HPO and other optimization algorithms as a control group.The results show that the optimization effect of the combination algorithm is better than that of the single algorithm optimization,and the optimization effect of the genetic-simulated annealing combination algorithm(GA-SA)is more advantageous in terms of accuracy and degree of fit,with an R 2 of 0.9375 for the test set,an MSE value of 0.0012,and a fitness function f(x)value of 0.0724.The model has better prediction performance and is more objective and intelligent compared to the original method.

关 键 词：输电边坡 危险性预测 支持向量回归 模拟退火算法 组合算法 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程] TN911-34[自动化与计算机技术—控制科学与工程]