Coordination of PV Smart Inverters for Grid Voltage Regulation

• Main Authors: Latani, T., Parameswaran, G., Priyanthan, G., Hemapala, K.T.M.U.
• Format: Conference Proceeding
• Language: English
• Subjects: Deep Deterministic Policy Gradient; Deep learning; Deep Reinforcement learning; Inverters; photovoltaic; Photovoltaic systems; Reactive power; Reinforcement learning; Silicon; Smart inverters; Training; Voltage regulation and PV curtailment
• ISSN: 2691-364X
• DOI: 10.1109/MERCon60487.2023.10355465

In the contemporary energy market, the utilization of photovoltaic (PV) is increasing considerably. This change brings new challenges to the power grid because of its variable and intermittent nature. One of the main issues is voltage violations and PV curtailment. A smart inverter (SI) provides a fast response method to regulate the voltage by varying real or reactive power at the point of common coupling (PCC). When multiple SIs operate under an autonomous control scheme, the reactive power level exceeds the threshold level. This creates an undesirable situation in the system. This paper mainly considers the coordination of the SI using a deep reinforcement learning algorithm (DRL). The DRL agent learns the policy through interaction with the IEEE-37 test feeder in the OpenDSS simulation to find out the optimal action. By defining the rewards scheme of the action carefully, the reactive power of SI can be utilized optimally, and the PV voltage will be maintained within the normal operating zone. Validation of the DRL agent's performance is done with the local autonomous control scheme. The results assure that a well-trained DRL agent can coordinate multiple SIs for voltage regulation and PV curtailment reduction.