基于MIMO-SVR的供热负荷日预报方法

Daily heat load forecasting method based on multi-input multi-output support vector regression

针对城市集中供热系统中提前24小时的日负荷预报方法具有较大误差问题,提出了一种基于多输入多输出支持向量回归(MIMO-SVR)的供热负荷日预报方法.该方法利用MIMO-SVR的多输出特性通过一步预报直接获得24小时的日负荷预报.通过对某热力站实际供热负荷数据进行仿真研究,结果表明,MIMO-SVR日预报的平均相对误差为2.47%,较多输入单输出支持向量回归(MISO-SVR)预报精度高,能够满足供热工程的应用需要.

In order to solve the large error problem of 24 hours advance daily heat load forecasting methods in city district heating system,a mthod of daily heat load forecasting method based on multi-input multi-output support vector regression （MIMO-SVR） was proposed. The method can directly obtain 24 hours daily heat load forecasting with the multi-output charactreristic of MIMO-SVR. The practical heat load data taken from a heating supply substation was simulated. The simulaion results show that the mean relative error of MIMO-SVR based daily forecasting is 2. 47% . The MIMO-SVR based forecasting method can give the higher precision compared with the forecasting method based on the multi-input single-output support vector regression （MISO-SVR）,and can meet the application demands in heat supply engineering.