Pipeline Network Decomposition for Parallel Optimization
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Large pipeline networks with hundreds of control devices pose computational challenges due to the complexity of optimizing control device settings, making it infeasible for existing computing systems to determine optimal settings within hardware and software constraints.
Innovation Solution
The approach involves decomposing the pipeline network into non-overlapping subnetworks, allowing for individual optimization of each subnetwork and parallel execution, which enables the computing system to optimize large-scale pipeline networks by transforming a serial problem into a parallelizable one.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If a network solver iterates through a search space performing a network simulation for each possible solution to determine optimal control device settings, then optimization accuracy is improved, but computational complexity and execution time increase to the point of infeasibility for large networks
Solution Approach 1:
The pipeline network is decomposed into multiple subnetworks based on control device groups, where each subnetwork contains a subset of control devices and associated pipeline segments. This segmentation allows the optimization problem to be divided into smaller, more manageable sub-problems that can be solved independently and in parallel, reducing overall computational complexity while maintaining optimization accuracy through coordinated boundary conditions.
2Productivity
If the pipeline network is decomposed into subnetworks for parallel optimization, then computational efficiency and productivity are improved, but solution convergence and optimality may be compromised
Solution Approach 1:
Boundary conditions at the interfaces between subnetworks are used to establish feedback loops where optimization results from one subnetwork influence the optimization of adjacent subnetworks. The system iteratively adjusts boundary conditions and re-optimizes subnetworks until convergence criteria are met, ensuring that the parallel decomposition approach yields a globally optimal or near-optimal solution while maintaining computational efficiency.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
A physical pipeline network is decomposed into multiple subnetworks. The subnetworks include upstream subnetworks and at least one downstream subnetwork. A network solver is executed on the upstream subnetworks in parallel to obtain a set of boundary conditions and a set of control device settings. The set of boundary conditions and a set of control device settings are then used to execute the network solver on the downstream subnetwork and obtain a result having another set of control device settings. The network solver may repeat executions until convergence is achieved. When convergence is achieved, the result is presented.