Digital twin for natural gas infrastructure operation and management via streaming dynamic mode decomposition with control

The gradual energy consumption increments and demand patterns diversification require an advanced operational protocol for natural gas infrastructure. However, conventional numerical methods suffer from low reliability and practicality due to the lack of usage of data that possess coherent physics in the system. The requirements to address this issue are system identification of data sequence and continuous modification to capture the changes of various environmental variables. Thus, we present the first data-driven digital twin model based on dynamic mode decomposition (DMD), which is well-suited for processing nonlinear, high -dimensional sequential data. It predicts pipeline flows and operating costs from measured data. Additionally, it adapts to changes in physical characteristics in pipelines. The proposed model was tested on real pipelines with a length of 5,000 km and demonstrated high accuracy, with a maximum error less than 0.2 MPa, and efficient computation, with a cost less than 2 s for 240 h. The proposed method constructs a reliable digital twin because it has the potential to be aware of the latent behavior of coherent physics in pipelines. The resulting digital twin forecast future status accurately and, based on this, may help establish advanced operating strategies in terms of safety and cost efficiency.