注意力机制和多目标粒子群混合驱动的配电网无功优化

Hybrid-Driven Reactive Power Optimization for Distribution Networks Driven by Attentional Mechanisms and Multi-Objective Particle Swarms

配电网安全稳定运行的关键之一是解决好配电网的无功功率平衡,尤其是高比例分布式光伏无序接入条件下,挑战巨大。为此,提出了一种注意力机制和多目标粒子群混合驱动的高比例分布式光伏配电网无功优化方法。首先,针对光伏出力强不确定性而预测难的问题,提出了基于Transformer的光伏预测模型,并与长短期记忆网络(long short-term memory,LSTM)以及卷积神经网络(convolutional neural networks,CNN)神经网络进行对比,验证了所提模型的精确性。进一步,针对强间歇性的分布式光伏高比例接入配电网的无功优化难题,提出了一种基于注意力机制的多目标粒子群算法(multi-objective particle swarm optimization algorithm based on attention mechanism,AMOPSO),将注意力机制应用于多目标粒子群算法,从而可减少粒子群算法本身不必要的搜索范围,有效地避免陷入局部最优,获得高比例分布式光伏配电网无功优化的调度策略。最后,在改进的IEEE33节点配电网系统中进行算例分析,验证了所提算法的有效性。

One of the keys to the safe and stable operation of distribution networks is to solve the reactive power balance,especially under the conditions of high proportion distributed photovoltaic disorderly access.Therefore,this article first addresses the strong uncertainty in photovoltaic output and difficulty in prediction.A Transformer based photovoltaic prediction model is proposed and compared with long short-term memory(LSTM)and convolutional neural networks(CNN)to verify the accuracy of the proposed model.Furthermore,a multiobjective particle swarm optimization algorithm based on attention mechanism(AMOPSO)is proposed to deal with difficulties in reactive power optimization in high proportion distributed photovoltaic distribution networks.The attention mechanism is applied to the multi-objective particle swarm algorithm,which can reduce the unnecessary search range of the particle swarm algorithm itself and effectively avoid falling into local optimization.Finally,a numerical example analysis is conducted in the improved IEEE33 node distribution network system,and the effectiveness of the proposed algorithm is verified through comparative analysis with particle swarm optimization and multi-objective particle swarm optimization.