Settling Time Sequence Forecast Method Research

The settling time sequence forecast method are very many, mainly divides into two big kinds, a kind is the theory method, a kind for basis actual material calculation subsidence quantity and time relations forecast method. Are very many as a result of the influence subsidence factor, but each factor the influence displays to the subsidence in the subsidence data, therefore the second kind of method obtained the widespread application. The grey systems theory method and the artificial neural networks method is in particular excavates the system interior through data itself the rule, obtains the forecast night-watch for is reasonable. This article the question which existed in the forecast has carried on the discussion on the grey systems theory and the artificial neural networks method.When grey systematic theory at present set up GM (1 , 1 ), it assumes that fit curve passes the first point of modeling data to confirm the integral constant, thus obtained to forecast formula. But as time goes on, the amount of information included of the first point of modeling data has been already insufficient. This text will confirm the constant method of integral divides into three kinds of schemes: (1)Pass the first point; ?Pass the last point;(3)Pass the point of smallest with the error. This text forecast 6 steps to 24 group data and compare of the result, have put forward it according to the method of choosing the scheme of curve type of the modeling data. Has carried on smoothness examination in the forecast to 24 groups of data, regarding does not satisfy smoothness condition 4 groups of data to use the accumulation to subside establishes the gray system forecast model, the forecast effect remarkable enhancement.While utilizing the neural network to forecast at present, adopt to subside totally as training samples mainly, this text will train the choosing of the sample to divide into twokinds of schemes: (1)Subside totally; (2)The interval subsides. This text forecast to same 24 group data and compare of the result, have put forward that chose different training sample according to different periods of subsiding.