Research On Distributed Voltage Control Of Distribution Network Based On Deep Reinforcement Learning

As the aim of energy strategy "carbon was at peak,carbon neutral" presented,the proportion of new energy represented by distributed photovoltaic(PV)in distribution network is increasing.In addition to improving energy utilization rate and reducing environmental burden,the peak and intermittency of new energy power generation will lead to the problem of voltage exceeding limit in time in power distribution network,which will harm the safe and stable operation of power distribution network.On the one hand,the traditional decentralized voltage control method will lead to different voltage regulating devices are difficult to coordinate,which can not get the global optimal voltage control strategy,and it is difficult to solve the decentralized voltage limit problem of the whole network;On the other hand,the large-scale application of distributed generations and the popularity of power electronic devices increase the uncertainty of power flow of the system,and the traditional voltage control algorithm which relies on power flow calculation has serious shortcomings in real-time voltage control.At the same time,how to meet the different response speed of different equipment when regulating voltage is also the focus of voltage control research.Thus,how to realize the intelligent coordinated regulation of different voltage regulating devices is a great challenge to solve the problem of voltage out-of-limit in distribution network.In view of the above problems,multi-time scale coordination voltage control method in regional distribution network and source-network-load-storage collaborative voltage method in distribution network with high permeability distributed photovoltaic are proposed.The main contributions of this thesis are as follows:1.Multi-time scale voltage coordination control strategy of regional distribution network: In order to solve the problem of local voltage out-of-limit,a multi-time scale voltage coordination control strategy is proposed considering the fast regulation of distributed photovoltaic output and the slow regulation of reactive power compensation of capacitor bank from the perspective of source and load.Firstly,the compensation position and compensation capacity of the capacitor bank were determined based on voltage sensitivity analysis,and the influence of photovoltaic power on node voltage was fully considered.Secondly,the output characteristics of photovoltaic inverter are analyzed,and the power output of photovoltaic and capacitor bank are adjusted respectively according to the defined control sequence.When the nodes are under voltage,only reactive regulation is used for photovoltaic,and when the nodes are over voltage,it observes the principle of that reactive power takes precedence over active power reduction to regulate voltage.Finally,deep reinforcement learning algorithm is used to solve the voltage control model,which can quickly obtain the optimal strategy of regulating power and effectively solve the problem of off-limit voltage in real time.2.Multi-terminal collaborative voltage control of distribution network with high-permeability distributed photovoltaic: Firstly,considering the output characteristics of different voltage regulating devices at multiple terminals of source-grid-load,the centralized coordinated and distributed coordinated voltage control architecture is established.Secondly,in order to solve the problem that the voltage in different periods in global distribution network,OLTC is used to adjust the standard voltage of the global network by changing the tap position to realize the regulation of whole feeder.Finally,in order to solve the problem of local voltage out-of-limit,a multi-terminal distributed cooperative voltage control method was proposed under the premise of dividing distribution network control area,and an optimization model was established to minimize the local voltage over-limit.Deep reinforcement learning algorithm was used to solve the problem,which ensured the effectiveness and speed of the regulation of the local voltage out-of-limit in the distribution network.The intelligent cooperation of different regulating devices is realized.