Transition towards a sustainable power system: MA-DA&DC framework based voltage control in high PV penetration networks

Abstract

Grid control systems must deal with voltage fluctuation and power losses because of the rising saturation of Renewable Energy Resources (RERs) in the Power Distribution Network (PDN). There have been several methods used to mitigate these inevitable voltage instabilities, but PV inverters are regarded as the most efficient ones. On the other hand, achieving optimum management and real-time coordination of several PV inverters in a PDN remains essential while preventing large power losses. The suggested work utilizes Decentralized Actors and Distributed Critics (DA&DC) architecture for multi-agents to obtain the optimal control and coordination of the agents (PV-inverters) incorporated into the distribution network. The critics employ a distributed control system, while the actors operate in a decentralized method. The proposed technology reduces computing complexity through multi-agent distributed training and decentralized execution. The suggested model is tested on IEEE-33, and results are obtained for the winter and summer seasons to analyze the influence of seasonal fluctuations on the outcomes of voltage variation, Power Losses (PL), Active Power (AP), and Reactive Power (ReP) of buses and PV systems. By reducing the voltage fluctuation ratio to 0.01327, the suggested approach achieves 85% voltage control. Moreover, average voltages of 0.9854 p.u. and 0.9950 p.u. were recorded throughout the summer and winter seasons, respectively. The PL are also maintained to an absolute minimum i.e., 0.0476 MW in the winter and 0.0535 MW in the summer. The results conclusively demonstrate that the proposed methodology outclasses its counterpart in both the summer and winter seasons.

Keywords

• Renewable energy sources
• Photovoltaic
• Reinforcement learning
• Distributed deep deterministic policy gradient
• Power distribution network