Study On The Prediction Model Of Natural Gas Hydrate Forming Condition

Natural gas hydrate is an efficient and clean resource that contains lots of heat energy.The reserves of natral gas hydrates are huge and has great strategic value.Since the discovery of natural gas hydrate,the governments of each country have always paid attention to natural gas hydrates.According to the requirements of the The 19th National Congress of CPC,the Central Committee of the Communist Party of China explicitly proposed to accelerate the industrialization of hydrate.In November 3,2017,the State Council agreed that natural gas hydrate has become the 173rd mineral species,thus laying an important role in the exploration and development of natural gas hydrate.The research,development and utilization of gas hydrate are beneficial to alleviate the contradiction between China's development and energy demand,promote the healthy development of the economy and improve the living standard of the people.In the process of pipeline transportation,natural gas is in high pressure and low temperature environment,and hydrate is easy to form.Once the hydrate is formed,it will gather in the bottom of the pipeline,reduce the circulation of the pipe,or even block the wellbore and lead to shut-in,which brings serious security risks and a large number of economic losses.In order to avoid the risk of hydrate formation in natural gas development,it is very necessary to predict the formation conditions of natural gas hydrate.This paper combines the Whale Optimization Algorithm(WOA)with the support vector machine(SVM)to establish a new method for predicting the formation conditions of natural gas hydrate in pure component,conventional gas,acid gas and alcohol salt system,as well as,using the grey correlation analysis method to determine the model input,and evaluating the model by using the anomaly detection method.The prediction accuracy of the Du-Guo model is the highest(AARD=0.32%),the correlation coefficient is the maximum(R2=0.9959),the prediction accuracy of the WOA-SVM model is the second(AARD =0.39%,R2=0.9795),and the Chen-Guo model has the lowest prediction accuracy(AARD =1.35%,R2=0.8832).By using the traditional model and intelligent algorithm model for pure components of natural gas hydrate calculation result,it is found that the correlation coefficient R2 of the training set of WOA-SVM model is 0.9979,the correlation coefficient of the prediction set is 0.9920,and the total AARD of the training and prediction is 2.86%.By the comparison of the 5 systems in the paper,we can see that the prediction error of all models increases in general when H2S content increases gradually.Among them,the vdW-P model(AARD=16.77%)has the lowest accuracy and the WOA-SVM model(AARD=2.66%)has the highest accuracy.When the temperature is low,the traditional thermodynamic model has a good accuracy.As the temperature rises,the error of the thermodynamic model increases gradually,while the prediction error of the WOA-SVM model rises slightly,which shows that the WOA-SVM model has a better adaptability and stronger stability than the traditional model in predicting the formation conditions of natural gas containing acid gas.Finally,all the data points in the text are detected based on the lever theory.The results show that there are 5 anomaly points in the Du-Guo model.The Parrish-Prausnitz model and the WOA-SAVM model only have an anomaly point,which indicates that the WOA-SVM model is more stable.It can be used as a new method for predicting the formation conditions of gas hydrate.