LS-SVM在城市供热系统热负荷预测中的应用研究

为了对热负荷及时准确的预测,采用最小二乘支持向量机(Least squares support vector machines,LSSVM)算法,结合网格搜索的交叉验证参数寻优建立预测模型。实验表明,与遗传算法参数寻优的SVM相比,计算速度提高27倍,均方误差提高3倍,拟合相关参数达到99%,说明该模型能快速准确的预测预测下一个工作日的短期热负荷,是一种可行的、有效的预测方法。

LCVR Phase Retardation Characteristic Calibration Method Using the LSSVM Model

In order to achieve the accurate prediction of heating load,Least squares support vector machine (LSSVM) is used, and the cross validation parameters optimization based on grid search is build to predict the model. Test re-sults show,compared with Genetic algorithm optimized parameters SVM,For prediction model based on LSSVM,the calculation speed is improved by 27 times, the mean square error is increased by 3 times, Squared correlation coeffi-cient reached 99%. The model could be rapidly and accurately obtained the short-term heat load of the next work day, prediction model based on LSSVM was an effective.