Research on cooperative control of distributed microgrid groups based on improved lookahead algorithm

Scale distributed power single machine into the network makes the power system inertia weakening, the frequency problem is more and more serious, in order to improve the reliability of the distribution network, can support each other distributed multi-area microgrid group came into being. However, the topology and operating environment of the microgrid cluster are more complex, in order to realize the problem of multi-region cooperative control of microgrid cluster, this paper proposes a novel control algorithm, i.e., the improved lookahead algorithm. Compared with the traditional reinforcement learning, the maximum expectation estimation represented by Q-learning overuses the greedy strategy, which in turn results in the overestimation of the action value; while the dual estimation method represented by DQ-learning uses too conservative a strategy and always abandons the greedy strategy in strategy selection, resulting in the underestimation of the action value in the process of strategy exploration. The interleaved Q-learning (IQ) algorithm uses the coupled estimation method to avoid the overestimation problem arising from the maximum estimator (ME) and the underestimation problem arising from the double estimator (DE). In this paper, based on Q-learning, interleaved Q-learning (IQ) is integrated into the lookaheadQ algorithm to design an improved lookahead algorithm, and at the same time, the qualification traces are integrated into the improved lookahead algorithm for algorithmic backtracking to obtain the maximum return in the past and the future.