基于深度学习的概率性电网潮流快速计算方法

A fast computing method of probabilistic power flow based on deep learning

高压直流输电技术能将大规模的清洁能源远距离输送,同时电网没有同步稳定问题和无功电压调节困难的约束,因此在全世界范围内具有广泛的发展前景。概率性电力潮流(PPF)在电力系统分析中起着至关重要的作用。但是由于电网潮流在传输中由于大的计算量,而导致消耗较大。传统的电网潮流计算方法对环境要求较高,时而不稳定,给PPF的实现带来了很大挑战。提出了一种基于模型的深度学习方法来克服计算的挑战。采用深度神经网络(DNN)逼近电力潮流计算,根据物理潮流方程进行训练,来提高学习能力。将分支流作为惩罚项加入到DNN的目标函数中,提高了DNN的逼近精度,简化了反向传播过程中使用的梯度,加快了训练速度,提高了算法的收敛速度。仿真结果验证了该方法的有效性和准确性。

HVDC transmission technology can transmit large scale clean energy over long distance,and the power grid is not constrained by synchronous stability problem and reactive voltage regulation difficulty.Therefore,it has a wide development prospect in the world.Probabilistic power flow(PPF)plays an important role in power system analysis.However,due to the large amount of calculation in the transmission of power grid,the consumption is relatively large.The traditional power flow calculation method has high requirements for the environment and is sometimes unstable,which brings great challenges to the realization of PPF.This paper proposes a model-based deep learning method to overcome the computational challenges.The deep neural network(DNN)is used to calculate the power flow and train data set according to the physical power flow equation to improve the learning ability.The branch flow is added to the objective function of DNN as a penalty term,which improves the approximation accuracy of DNN,simplifies the gradient used in the process of back propagation,speeds up the training speed and improves the convergence speed of the algorithm.Simulation results show the effectiveness and accuracy of the proposed method.