A Data-Driven Genetic Algorithm for Power Flow Optimization in the Power System With Phase Shifting Transformer

Abstract

Phase-shifting transformer (PST) is one of the flexible AC transmission technologies to solve the problem of uneven power transmission. Considering that PST can also be used as a regulation means for the economic operation of the system, it is necessary to study the power flow optimization of power systems with PST. In order to find a more efficient power flow optimization method, an improved genetic algorithm including a data-driven module is proposed. This method uses the deep belief network (DBN) to train the sample set of the power flow and obtains a high-precision proxy model. Then, the calculation of the DBN model replaces the traditional adaptation function calculation link which is very time-consuming due to a great quantity of AC power flow solution work. In addition, the sectional power flow reversal elimination mechanism in the genetic algorithm is introduced and appropriately co-designed with DBN to avoid an unreasonable power flow distribution of the grid section with PST. Finally, by comparing with the traditional model-driven genetic algorithm and traditional mathematical programming method, the feasibility and the validity of the method proposed in this paper are verified on the IEEE 39-node system.