Water quality predication method based on grey theory and support vector machine

Abstract

The invention discloses a water quality predication method based on a grey theory and a support vector machine and belongs to the technical field of water quality predication methods. The method comprises the following steps: (1) carrying out wavelet decomposition on an acquired data set; (2) carrying out translation on detail sub-sequences decomposed by the step (1) and carrying out phase-space reconstruction; (3) establishing a predication model of each detail sub-sequence according to each detail sub-sequence subjected to the phase-space reconstruction in the step (2); (4) establishing a scale sub-sequence predication model according to scale sub-sequences decomposed by the step (1); (5) inputting data and predicating, namely carrying out reverse translation on a predicated value of each obtained detail sub-sequence, and reconstructing the predicated values of the scale sub-sequence predication models and the detail sub-sequence predication models to obtain a final predicated result. Compared with a traditional water quality predication method, the method has high predication precision and a wide application range. Meanwhile, the water quality predication method has good robustness and complex nonlinear mapping capabilities in the aspects of predication stability and precision.

Description

technical field [0001] The invention relates to a water quality prediction method, in particular to a water quality prediction method based on gray theory and support vector machine. Background technique [0002] Water quality prediction is an important method for water resource management and water pollution control, and is the basic work for water quality control and water resource development and utilization. basis and technical support. There are many methods for water quality prediction, typical ones are based on mathematical statistics, chaos theory and neural network methods. Mathematical statistics method has the problem of difficulty in predicting multi-factors of water quality. The application of chaos theory requires a large amount of water quality information. The neural network prediction method can describe the nonlinear characteristics of complex changes in water quality time series, but it has a slow network convergence speed. , It is easy to fall into prob...

AI Technical Summary

Problems solved by technology

However, there are still some problems in the prediction of water quality time series using gray theory and support vector machine: First, the fitting prediction accuracy of support vector machine to water quality time series is affected by its parameter selection. The optimal method is limited by the condition that there is no interaction between various factors, which is not only very time-consuming, but also often cannot guarantee that the obtained parameters are optimal; second, the time series changes of water quality monitoring parameters are affected by many factors. Some monitoring values ​​show irregular changes such as sudden jumps, resulting in the fact that the ideal model is often not obtained during the model training process; third, although the wavelet decomposition method can obtain a stable subsequence, it can reduce the The impact of data on the prediction model, but a large amount of input training data will still affect the prediction accuracy of the model

Images