Research On Production Allocation Of Oilfiled Based On Multi-objective Optimization

At present,most of the oil fields in my country are in the middle and later stages of development.Along with the development,a series of problems have appeared in the oil field: the oil well has a high degree of recovery,is in the stage of high water cut and extra high water cut for a long time,and it is difficult to develop;the cost of crude oil in various oil extraction plants rises rapidly,and the cost distribution is uneven;the capital investment and turnover is difficult.These problems have led to a gradual decline in oil production and reduced efficiency.Under this background,how to use machine learning methods on the basis of the existing resources and technologies for the petroleum enterprises,improve the level of oil field production forecast,construct a reasonable multi-objective optimization model of oil field production proration,achieve oil field production and financial support of scientific planning,improve the level of petroleum enterprise decision-making,this will be the major strategic objectives for the sustainable development of oil field company.(1)Aiming at the problem of low prediction accuracy of traditional prediction models and incomplete consideration of influencing factors,are not comprehensively considered,this paper proposes a prediction method based on ARIMA-LightGBM-LSTM model fusion.First,use the ARIMA model to make univariate predictions on the output column,put the predicted values as new variables and the features obtained by feature selection into the LightGBM model for mining different attributes,and merge the predicted values into the second feature selection of multivariable sequences.The resulting multivariate sequences are then predicted using the LSTM model.The final prediction value is calculated by weighting and combining the prediction results of the three models,and obtaining the best weights through multiple experiments.The simulation data results show that the multivariate prediction method based on ARIMA-LightGBM-LSTM weighted combination can obtain more accurate prediction values than a single prediction method.(2)The MOEA/D algorithm is improved to solve the production allocation problem.Oilfield allocation is a multi-objective optimization problem of rational allocation of output,investment capital and output profit.The three optimization goals are the largest total output,the largest total capital investment and the largest total profit.This paper is based on the differential evolution algorithm based on decomposition,the Q matrix memory algorithm is introduced.And a MOEA / D-Q algorithm that is more suitable for solving multi-objective optimization problems is designed to complete the model.In the research process,an oilfield was used as an example to conduct experiments,and the oilfield allocation plan for 2009 was designed based on the multi-objective optimization results.Compared with the real oilfield production matching results,it is proved that the algorithm can achieve good results in solving multi-objective optimization problems.