基于贪婪策略的多层自动发电控制

Multiple Level Automatic Generation Control Based on the Greedy Strategy

分布式能源大规模并网所带来的随机扰动问题,影响电力系统的安全稳定和经济运行。该文提出一种将具有多步前瞻属性的贪婪控制算法与基于两层功率分配模式的功率优化分配算法相结合的多层自动发电控制策略(multiple level automatic generation control,ML-AGC),以解决分布式能源大规模并网所带来的随机扰动问题。即在ML-AGC的控制部分采用具有快速收敛特性的多步前瞻贪婪迭代算法,来获取自动发电控制系统的总功率指令;分配部分采用具有交互协同和自学习特点的协同一致性分层Q学习(hierarchical Q-learning based collaborative consensus,HQCC)算法,来提升功率分配策略在强随机扰动下的适应性。通过对改进的IEEE两区域负荷频率控制模型,以及融入大量分布式能源及冷热电联产的智能配电网模型进行仿真验证,结果表明所提策略能够获取多区域的最优协同控制及机组功率的动态优化分配,能够解决分布式能源大规模并网所带来的强随机扰动问题。与多种方法相比,ML-AGC具有更快的动态优化速度和更低的发电成本。

The stochastic disturbance caused by large-scale grid connection of distributed energy affects the security, stability and economic operation of power system. In this paper, a multiple level automatic generation control(ML-AGC) strategy was proposed, which combined the greedy control algorithm with multiple step look-ahead attribute and the power optimal allocation algorithm based on two-level power allocation mode to solve the strong stochastic disturbance caused by large-scale grid connection of distributed energy. In the control part of the proposed strategy, the multiple step look-ahead greedy iterative algorithm with fast convergence was used to obtain the total power commands of the automatic generation control(AGC) system;at the same time, the hierarchical Q-learning based collaborative consensus(HQCC) algorithm with interactive collaboration and self-learning characteristics was used in the allocation part to improve the adaptability of the power allocation strategy under stochastic disturbance.The simulations of the improved IEEE two area load frequency control power system model and the smart distribution network model with a large number of distributed energy and combined cooling heating and power show that the proposed strategy can achieve the optimal coordinated control and power allocation of the power system. Compared with many algorithms, ML-AGC has faster dynamic optimization speed, which can reduce the cost of power generation.