Deep Neural Networks for gas transmission network
MP DATA was commissioned to develop and industrialize an innovative solution to track gas quality throughout a national transportation network. The developed solution was based on a hybrid model applying Machine Learning and Finite Element Analysis principles.
Following a state-of-the-art study, we have designed a modular and hybrid simulation solution, thus, exploiting a series of local models capable of adapting to any topology in real time. The compute engine was at first developed in Python programming language and then optimized at maximum capacity using the Cython programming language. Once the real-time simulation was designed, we integrated numerous additional modules (slow leaks detection and targeting, data reconciliation in the event of faulty measurements, reconstitution of stocks in line, etc.). Finally in terms of industrialization, the resulting solution was deployed and integrated in the client’s AWS cloud environment.
Precision and adaptability of the solution
Reducing the simulation error in the event of measurement failure, thanks to the reconciliation of the measurements
Data pipeline mastery
By using a turnkey solution offering connectivity to third-party cloud solutions