Loading [MathJax]/extensions/MathMenu.js
Topology-Aware Learning Assisted Branch and Ramp Constraints Screening for Dynamic Economic Dispatch | IEEE Journals & Magazine | IEEE Xplore

Topology-Aware Learning Assisted Branch and Ramp Constraints Screening for Dynamic Economic Dispatch


Abstract:

Multi-interval or dynamic economic dispatch (D-ED) is the core of various power system management functions. This optimization problem contains many constraints, a small ...Show More

Abstract:

Multi-interval or dynamic economic dispatch (D-ED) is the core of various power system management functions. This optimization problem contains many constraints, a small subset of which is sufficient to enclose the D-ED feasible region. This paper presents a topology-aware learning-aided iterative constraint screening algorithm to identify a feasibility outlining subset of network and generating units ramp up/down constraints and create a truncated D-ED problem. We create a colorful image from nodal demand, thermal unit generation cost, and network topology information. Convolutional neural networks are trained for constraint status identification using colorful images corresponding to system operating conditions and transfer learning. Filtering inactive line flow and ramp up/down constraints reduces the optimization problem's size and computational burden, resulting in a reduction in solution time and memory usage. Dropping all inactive branch and ramp constraints may activate some of these originally inactive constraints upon solving the truncated D-ED. A loop is added to form a constraints coefficient matrix iteratively during training dataset preparation and algorithm utilization. This iterative loop guarantees truncated D-ED results feasibility and optimality. Numerical results show the proposed algorithm's effectiveness in constraint status prediction and reducing D-ED size and solution time.
Published in: IEEE Transactions on Power Systems ( Volume: 37, Issue: 5, September 2022)
Page(s): 3495 - 3505
Date of Publication: 13 January 2022

ISSN Information:

Funding Agency:


References

References is not available for this document.