Reliability assessment of underground gas storage units based on BP neural network and Monte Carlo simulation

Underground gas storage(UGS)is one of the most important means of natural gas peak shaving, gas pipeline network balancing and strategic energy reserving.The quantitative evaluation on the reliability of UGS gas injection and production process is beneficial to UGS operation management and its natural gas supply capacity assessment.In this paper, the fluctuation of geometrical dimensions and operating parameters of each unit was characterized by means of Monte Carlo sampling method based on its operating conditions and running time in the actual operation process of UGS.Then the reliability of the key units which have influence on UGS gas injection/production rate was simulated.To avoid the impact of insufficient simulation times on the calculation result and reduce the computation time of unit reliability model, the BP neural network algorithm was imported in the UGS unit calculation model.And then, the unit reliability model was built up by using the BP neural network algorithm.In this model, UGS process parameters and operation time are taken as the input variables and the unit reliability which is calculated by means of Monte Carlo Simulation is taken as the output value.Finally, in order to verify the validity of this method, the operation reliability of key equipment units in one certain UGS was calculated based on the operational conditions of the UGS and the basic parameters of its unit equipments.