作　　者： ; ; ; ; (代州）; (傅闯）;

机构地区： 华南理工大学电力学院

出　　处： 《高电压技术》 2013年第11期2822-2828,共7页

摘　　要： 为对输电线路覆冰进行有效地监测、预测及预警,提出了一种基于支持向量机(support vector machine,SVM)的输电线路覆冰回归模型,用于输电线路覆冰情况的短期预测。这一研究工作是在MATLAB环境下,应用LIBSVM软件包编程进行建模仿真的;针对实测微气象-覆冰数据多维、自由度大的特性,选定与覆冰相关性最大的气温、相对空气湿度数据以及覆冰参考量作为输入量,覆冰质量作为输出量;提出了基于支持向量机的超短期预测、短期迟滞预测和滚动预测3种预测模型,并通过实例数据仿真评估了模型的有效性。结果表明:超短期预测模型预测精度>90%,但时效仅15min、实用价值较低;短期迟滞预测模型和滚动预测模型在2h内预测精度均>80%,可适用于输电线路覆冰的短期实时预测;滚动预测模型理论上可预测更长期的覆冰情况,假设微气象参量恒定不变限制了其预测精度,若结合微气象预报将会有更好的预测效果。由于目前适用于建模仿真的完整覆冰数据较少,因此支持向量机用于建立输电线路覆冰回归模型的有效性和稳定性还有待进一步验证。 For effectively monitoring and forecasting icing on transmission lines, we proposed a regression model of transmission line icing based on support vector machine（SVM） for short-term forecasts, and performed the simulations using the LIBSVM software package in MATLAB. In these simulations, according to the measured micro weatheR-icing data, which are multi-dimensional and have large degree of freedom, we selected three parameters, i.e. temperature, relative air humidity, and reference icing data, as the input variables, and icing weight as the output variable. Moreover, we proposed three models based on $VM, namely, the super short-term forecast model, the short-term hysteresis forecast model, and the rolling forecast model, and then assessed the efficiency of simulations using practical data. The results show that the super short term forecast model has prediction accuracy higher than 900/00 in 15 minutes duration, which is quite short and makes the model far from being applicable. Accuracy of the short-term hysteresis prediction model and the rolling forecast model are both higher than 80% in 2 hours, so these two could be applied to short-term prediction of transmission line icing. Moreover, using the rolling forecast model can theoretically predict long-time icing, but the assumption of constant micro meteorological parameters may limit its prediction accuracy, and a better prediction can be achieved by combining with micro weather forecasts. At present, only a few complete data suit modeling and simulation, so the correctness and stability of the proposed regression model need further verification.

关 键 词： 覆冰 输电线路 支持向量机 回归模型 短期预测 在线监测

领　　域： [电气工程]