Forecast And Application Of Fuzzy Time Series In Drilling Risk Probability

In recent years,it has attracted the attention of researchers at home and abroad about the risk probability prediction problem of drilling operation site.Its potential imprecision and fuzziness make some prediction methods no longer applicable.Fuzzy technology has become a useful tool,fuzzy time series method has also been generated.Most of the researchers have improved this method and applied it to many fields,and the effect is very significant.Based on fuzzy time series method,a new fuzzy time series method is proposed for drilling risk prediction.First,the division of the discourse domain and the introduction of the fuzzy set.Time series is tested by stationarity.Fuzzy C means clustering algorithm(FCM)is used to get the cluster center and membership matrix.Then the data set is fuzzy divided.The principle of division is to use the extremum of the sequence and the average value of the adjacent two cluster centers as the boundary value of the sub-interval,and introduce the fuzzy set.Second,modeling and prediction.According to the membership matrix,the sub membership sequence is obtained,the number of sequences and fuzzy sets are the same.The sub membership sequence is time series analysis to obtain the best prediction model,the analysis steps including stationarity test,white noise test,model identification and order determination,model parameter detection and residual test.According to the optimal model,it can predict the future value of the moment.Third,correction and defuzzification.After judging the membership of the data initially,the prediction value is revised according to the concept of membership degree.The correction principle use the min-max standardization to get the prediction value as far as possible,so as to make the membership more intuitive,clear and accurate.Secondly,the weighted average method of membership degree is used to defuzzification the corrected value,and the same type of prediction result with the original observation sample is obtained.Finally,the new method is compared with the method of Song and Chissom and the method of Tsaur and Kuo.The results show that Root mean square error(RMSE)and Mean absolute percentile error(MAPE)of the proposed methods are less than those of the other two prediction methods.Therefore,it is easy to get the new method in the field of drilling operation site risk probability prediction compared with the existing prediction method,which has a certain scientific significance.