Prediction of Thermal Recovery Casing Stress Based on Improved PSO-BP Neural Network
In the process of heavy oil thermal recovery,plastic deformation of the oil casing string under the action of multi-ple loads such as temperature and formation leads to fracture or failure.In this paper,according to the high temperature service condition of the thermal recovery pipe column,the asynchronous change learning factor and adaptive weights were introduced to establish an improved PSO-BP model with the input parameters of steam injection temperature,well depth,non-uniformity coeffi-cient and cement ring temperature,and the output parameter of casing stress.In the paper,taking N80 thermal casing as an example,the finite element simulation results at five temperature conditions,namely,260℃,280℃,300℃,320℃and 340℃,were selected as the training data to compare the predicted and experimental values and finite element calculated values of the casing stresses of the BP model,the GA-BP model,the MEA-BP model,the PSO-BP model,and the improved PSO-BP model at the well depths of 200 m,300 m,400 m,500 m,600 m,and 700 m at the temperature of 300℃working condition.The results show that the stresses predicted by the improved PSO-BP model are the closest to the test values,with the maximum and minimum errors of 2.69%and 0.06%,respectively.Finally,the full paper evaluates the established improved PSO-BP model in terms of training data,prediction error,and computation time,which provides an intelligent and efficient model for the strength and safety analysis of thermal casing columns in the service process.
BP neural networkstresspredictive modelparticle swarm optimisation algorithm
NETL
NSTL
万方数据
维普
