基于分块坐标下降思想的并行无功优化分解协调算法

A Block Coordinate Descent-Based Parallel Decomposition-Coordination Algorithm for Reactive Power Optimization

针对无功优化分解协调模型求解中增广拉格朗日函数的不可分问题,提出了基于分块坐标下降(block coordinatedescent,BCD)思想的并行分解协调计算方法。该方法可实现全网无功优化的分解协调计算,仅需要在相邻分区之间交换边界节点的功率和电压用于协调,解决了大规模电网集中式无功优化存在的计算速度慢和数据传输瓶颈问题;而且各控制中心可自主选择优化算法,实现了自律分散与协调控制的结合。算例结果表明,该算法可以大大减少全网无功优化的计算时间,并且与基于辅助问题原理的分解协调算法相比,其收敛速度更快、计算效率更高。

In allusion to inseparability of augmented Lagrangian function during the solution of decomposition-coordination mode for reactive power optimization model, a parallel decomposition-coordination algorithm based on the thought of block coordinate descent （BCD） is proposed. The proposed algorithm can implement decomposition and coordination of reactive power optimization of the whole grid, in which only power and voltage information of boundary nodes between adjacent sub-areas is needed to be exchanged for the coordination, thus both defects of slow computing speed and bottleneck of data transmission existing in centralized reactive power optimization of large-scale power grid are remedied. Besides, each sub-area needs not internal model and data of other sub-areas, thus each control center can choose optimization algorithm independently and the combination of autonomous decentralization with coordinated control. Simulation results of IEEE ll8-bus system and IEEE 300-bus system show that the proposed algorithm can speed up the computation of reactive power optimization of the whole power grid obviously, and comparing with the decomposition-coordination algorithm based on auxiliary problem principle （APP） the convergence rate of the proposed algorithm is faster and its computational efficiency is higher.