RAILWAY TRACTION SUBSTATION OPTIMIZATIONS WITH TABU SEARCH APPROACH

Abstract

This document reports on the research of traction substations optimization using Tabu Search strategy. In a DC railway system, traction substations convert power from AC to DC and supply trains moving along the line. For traction substation with inverter, it could also return train regenerative energy to the AC supply system. Operating cost of an electrified railway is mainly contributed by the energy cost. To reduce traction energy cost, one of the ways is to optimize the design of traction substations. Many aspects of traction substations can be optimized, including substation location, justification of inverter, substation size, transformer tap ratio and converter firing angle. The design of traction substations can be optimized with respect to various objectives. Normally, the objective functions involve simulating train operation for a period of time such as calculating traction energy consumption and average rectifying or inverting power. The evaluation of objective functions, and thus the optimization program, requires long computation time. A modem heuristic search strategy known as Tabu Search has been proposed for the optimization of traction substations. The optimization method has been integrated with a train simulator. Together, the program developed has been used to optimize traction substation location, inverter placement and rectifier firing angle. This has been verified with 2 objective criteria: traction energy consumption and equal load sharing. Tabu Search is found suitable for the optimization of traction substations. It is able to obtain satisfactory solution with small number of iterations. The flexibility in objective function requirement allows the evaluation function that includes train simulation. Besides, it provides generality to the optimization program such that change of objective is simply a switch of objective function. For time critical application, Tabu Search allows trade-off between computational time and optimality though it normally reaches near-optimal solution quite fast. Traction substation location has been optimized successfully. The study case results seem to suggest some patterns .associated with the objective functions. The traction energy minimization leads to sites of uniform and maximum separation. Meanwhile, the load sharing optimization gives rise to sites that are clustered. Since rectifier firing angle control is supposed to be an on-line operation with trains moving in real-time, it is noted that the model might not be comprehensive enough and computational time is too long. However, implementation is still possible if the model could be updated constantly with actual measurement and fast computing power is available. Traction energy is found not very sensitive to optimization. It has shown low improvement (about 0.7%) in the test cases. For load sharing problem, the objective function has tested to be more sensitive on optimization of traction substation location and rectifier firing angle. In addition, a substation outage test simulated has shown that the optimized configuration is able to make effective reduction of peak power flow at traction substations.