Utilization of Artificial Intelligence for Voltage and Reactive Power Control and Its Operational Results

The Nagano-region grid in Japan is a radial system transmitting large-scale power from major power plants over long distances. This configuration causes complex issues related to synchronous stability, frequency, and voltage under both normal and fault conditions. In recent years, the expansion of photovoltaic (PV) installations has intensified voltage fluctuations under normal conditions, requiring frequent voltage adjustments to maintain appropriate voltage levels. The conventional voltage and reactive power control (VQC) method adjusts generator terminal voltage based on predetermined voltage setpoints calculated for the most severe system conditions. Consequently, generator voltage control has been limited, and most adjustments have relied on reactive power compensation devices and transformer taps. This has led to increased operation counts for these devices, posing a risk of significantly higher maintenance costs in the future. While flexible adjustment of generator terminal voltage is an effective countermeasure, determining appropriate voltage values in real-time according to changing system conditions poses significant challenges for operators. To address this, we developed a new VQC method utilizing artificial intelligence (AI). The new VQC method consists of two components: off-line learning and on-line control. In off-line learning, historical grid data for one year is used to calculate the optimal control settings for equipment such as reactive power compensation devices, transformer taps, and generator terminal voltage. These settings are then used to train the AI models. In on-line control, real-time grid data is input into the trained AI model to output the optimal settings. The new VQC method has been operational since May 16, 2023. From the operational results, we confirmed that system operators can achieve optimal voltage control with the assistance of AI.