Optimization of gas storage based on improved multi-objective particle swarm optimization algorithm
In order to reduce the number of microseisms caused by unreasonable gas injection in the gas storage and ensure the maximum gas injection in the gas storage,a surrogate model based on bi-directional long short-term memory(BiLSTM)neural network was constructed to predict the number of microseisms and the effective stress in the gas storage,and the prediction error was within 5%.An elite evolutionary multi-objective particle swarm optimization(EMPSO)algorithm was proposed,which used sorting grouping strategy to group the population.Random elite competition learning was conducted within each group to improve the diversity of the algorithm,while introducing the idea of elite aggregation to accelerate the convergence speed of the algorithm.Based on the BiLSTM model and EMPSO algorithm,the gas injection process of gas storage was optimized,and compared with three other multi-objective optimization algorithms.The EMPSO algorithm was applied to actual production optimization.The results showed that the improved algorithm had better Pareto front and faster convergence speed.After optimization,the number of microseisms and effective stress were reduced by 9.78%and 10.12%,respectively.It was of great significance to ensure the safety of gas storage and improve the storage capacity of gas storage.
underground gas storagesurrogate modelbi-directional long and short-term memoryimproved particle swarm algorithmmulti-objective optimisation search
NETL
NSTL
万方数据
