Application Of Adaptive Neuro-Fuzzy Inference System (ANFIS) In Monthly Runoff Prediction

Runoff prediction is an important part of water resources analysis.However,due to the influence of climate change,geographical location,human activities and other factors,the change of runoff is complicated,which makes runoff prediction quite difficult.This article is based on the past 31 years runoff data of the Lanzhou Station,through the MATLAB programming to realize the Adaptive Network-based Fuzzy Inference System(ANFIS)in the application of monthly runoff prediction,and through the two pre-built ‘function' function(including the regression equation,the coefficient of determination,the correlation coefficient,etc.)to evaluate the results.It can be seen from the output results :(1)the overall results of ANFIS in the monthly runoff forecast of lanzhou station are satisfactory,and there are 60 groups in the test data,among them,the percentage of errors in 51 groups was less than 10%,and the results of 7 groups ranged from 10% to 20%,and the results between the two groups ranged from 20% to 30%;(2)The correlation coefficient(between 0 to 1,the higher the value is,the higher the correlation is)between the predicted value and the actual value both more than 0.8 in the 12 months;(3)In the output of the regression equation,in January,April,June results compared with the results of the remaining nine months have some disparities,the coefficient of determination(to judge the advantages and disadvantages of regression equation as a result,between 0 to 1,the higher the value is,the better the result of the regression equation)between 0.4 to 0.7,the remaining nine month coefficient of determination were greater than 0.75.When ANFIS is used to predict the monthly runoff in Lanzhou Station,the predicted value is accurate.More Than This,the model is also accurate in predicting the trend of monthly runoff,with high predictive power.According to the different requirements,it can change the network form or setting different parameters to adjust the model to achieve better results.It is a worth exploring method in runoff or hydrological forecasting.