现有深度强化学习(deep reinforcement learning,DRL)方法在解决配电网电压优化问题时,存在信用分配难、探索效率低等问题,在模型训练速度和优化效果等方面表现欠佳。为此,结合配电网分区降损与模仿学习的思想,提出一种基于指导信号的...现有深度强化学习(deep reinforcement learning,DRL)方法在解决配电网电压优化问题时,存在信用分配难、探索效率低等问题,在模型训练速度和优化效果等方面表现欠佳。为此,结合配电网分区降损与模仿学习的思想,提出一种基于指导信号的多智能体深度确定性策略梯度(guidance signal based multi-agent deep deterministic policy gradient,GS-MADDPG)的电压优化方法。首先,将电动汽车(electric vehicles,EV)集群、分布式电源(distributed generations,DG)和无功调节装置作为决策智能体,构建强化学习优化模型。然后,通过配电网分区,解耦多智能体的外部奖励,并结合模仿学习,利用指导信号引入内部奖励,帮助智能体快速寻优。最后,基于改进IEEE33节点系统进行算例测试。结果表明,所提电压优化策略较传统DRL方法具有更高的样本利用率,实现了更稳定的收敛及更高的模型训练效率,提升了配电网电压的优化效果。展开更多
Empirical and deterministic models have not proven to be effective in path loss predictions because of the problems of computational complexities, low accuracies, and inability to generalize. To solve these problems r...Empirical and deterministic models have not proven to be effective in path loss predictions because of the problems of computational complexities, low accuracies, and inability to generalize. To solve these problems relating to path loss predictions, this article presents an optimal path loss propagation model developed at 3.4 GHz with the use of fuzzy logic. We introduced Fuzzy logic to accurately represent all forms of uncertainties in the data spectrum as the signal propagates from the transceiver to the receiver, thereby producing accurate results. Experimental data were collected across Cyprus at 3.4 GHz and compared with three existing path loss models. The fuzzy-logic path loss prediction model was then developed and compared with the experimental data and with each of the theoretical empirical models, the newly developed model predicted signal loss with the greatest accuracy as it gives the lowest root-mean-square error. The newly developed model is very efficient for signal propagation and path loss prediction.展开更多
文摘现有深度强化学习(deep reinforcement learning,DRL)方法在解决配电网电压优化问题时,存在信用分配难、探索效率低等问题,在模型训练速度和优化效果等方面表现欠佳。为此,结合配电网分区降损与模仿学习的思想,提出一种基于指导信号的多智能体深度确定性策略梯度(guidance signal based multi-agent deep deterministic policy gradient,GS-MADDPG)的电压优化方法。首先,将电动汽车(electric vehicles,EV)集群、分布式电源(distributed generations,DG)和无功调节装置作为决策智能体,构建强化学习优化模型。然后,通过配电网分区,解耦多智能体的外部奖励,并结合模仿学习,利用指导信号引入内部奖励,帮助智能体快速寻优。最后,基于改进IEEE33节点系统进行算例测试。结果表明,所提电压优化策略较传统DRL方法具有更高的样本利用率,实现了更稳定的收敛及更高的模型训练效率,提升了配电网电压的优化效果。
文摘Empirical and deterministic models have not proven to be effective in path loss predictions because of the problems of computational complexities, low accuracies, and inability to generalize. To solve these problems relating to path loss predictions, this article presents an optimal path loss propagation model developed at 3.4 GHz with the use of fuzzy logic. We introduced Fuzzy logic to accurately represent all forms of uncertainties in the data spectrum as the signal propagates from the transceiver to the receiver, thereby producing accurate results. Experimental data were collected across Cyprus at 3.4 GHz and compared with three existing path loss models. The fuzzy-logic path loss prediction model was then developed and compared with the experimental data and with each of the theoretical empirical models, the newly developed model predicted signal loss with the greatest accuracy as it gives the lowest root-mean-square error. The newly developed model is very efficient for signal propagation and path loss prediction.
