Prediction And Analysis Of Jinan Rainfall Based On Weighted Markov Chain Combination Model

Rainfall refers to the depth of the rainwater that falls from the sky to the ground without evaporation,infiltration,and loss and accumulates on the water surface.Precipitation is an important basis for the calculation of regional water resources.Predicting rainfall can take preventive measures in advance,minimize losses,and provide decision-making basis for energy development time and location selection.Therefore,finding an accurate and simple method to forecast rainfall is necessary.This paper takes the rainfall of Jinan City,Shandong Province from 1958 to 2018 as the research object,with a total of 61 years of data as the training set,and the 2019-2020 data as the test set.This paper builds time series ARIMA model,simple exponential smoothing model,weighted Markov chain model to make predictions of rainfall.Since the Markov chain model has the ability to predict the future given the current knowledge or information,only the current state is used to predict the future,and the historical state is irrelevant for predicting the future,that is,the state has no aftereffect.This also leads to poor prediction of extreme values.Therefore,weighted Markov chain model has improved at last,and a grey prediction-weighted Markov chain combination model has established,which refers to the prediction error of the grey model is modeled by Markov model,and finally the predicted rainfall is obtained.At the last,the above four models are optimized to establish a combined model to obtain the predicted value of the combined model,and the prediction result is better than the original model,and the prediction result is improved.The main research work of this paper is as follows:(1)After the stationarity test of the sequence,the time series model is established,and the exponential smoothing image is fitted,and the ARIMA model parameters are selected through the autocorrelation function image and the BIC test image,and the ARIMA model is fitted.Finally,based on the two models,the rainfall in 2019 and 2020 is predicted,and the relative error is calculated.(2)By classifying the state of the rainfall sequence according to the mean-mean-square deviation method,calculating the transition frequency matrix and the state transition matrix,and performing the Markov test to establish a Markov model,obtain the prediction interval,and introduce the fuzzy set level eigenvalue to determine Predictive value.(3)Through the feasibility test of the rainfall sequence and the three-point moving average,an unbiased gray prediction model is established,the prediction error is clustered in an orderly manner to establish a Markov model,and a Markov test is performed to establish a Markov model,obtain the prediction interval,and introduce fuzzy set-level eigenvalues to determine the predicted value.(4)The fitting value of the above model is weighted by the improved combination model of the genetic algorithm to obtain a new fitting value,the fitting effect of all models is compared by calculating the cumulative error,and finally the combined model is used to predict the rainfall in 2019 and 2020.