基于BP神经网络与蒙特卡罗法的储气库单元可靠性评价

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

储气库是天然气用气调峰、平衡管网、能源战略储备的重要手段,对储气库注采气过程可靠性展开定量评价,有利于储气库运行管理、供气能力评估。以储气库实际运行过程中各单元的运行工况、运行时间为前提条件,采用蒙特卡罗抽样方法表征单元几何尺寸、运行工艺参数的波动情况,对影响储气库注采气量关键单元的可靠性进行建模。为了规避蒙特卡罗抽样次数不足带来的影响,同时缩短单元可靠性模型运算时间,在储气库单元计算模型中引入BP神经网络算法。以储气库工艺参数、运行时间为输入变量,以蒙特卡罗法计算单元的可靠度为输出值,应用BP神经网络算法建立单元可靠度模型。结合某储气库的工况条件、单元设备基础参数,对其中关键设备单元的运行可靠性进行计算,验证了该方法的有效性。

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.