用于过程控制的灰色预测模型性能测试

The Performance Tests for the Grey Forecasting Model Used in Process Control

针对过程控制应用背景,对灰色预测模型进行了多项性能测试。结果表明:灰色预测模型不适用于过零变化输入数据序列;对于不符合指数规律变化的信号虽能适用但有较大的预测误差;在固定采样周期下所能适用的输入信号的变化率有一个上限;当输入信号趋于不变时会出现被零除的数值不稳定情况;预测误差将随预测步长或建模维数或采样周期的增大而增大。

To the process control application,the performance tests for the grey forecasting model are carried out.The test results show that the grey forecasting model is not applicable to the input data with zero-crossing,and it causes a big predictive error at nonexponential input data,an upper limit of the input signal change velocity at a fixed sample time,and a situation of unstable numerical computation when the input variable is no change,and the predictive error increases in direct ratio with predictive step length or model dimension or sample time.