The amount of photovoltaic (PV) introduced in Japan is increasing year by year. On the other hand, the issue with PV is that the amount of power generated fluctuates due to changes in the weather, which can disrupt the power demand and supply balance and have negative effects on the power system. To address this issue, technology that accurately predicts PV power generation is essential. Therefore, we predicted total PV power generation for the next day in the Tokyo Electric Power Company (TEPCO) area. In this paper, we proposed a method that adds normalized images of variables forecasted from Meso‐Scale Model (MSM) of Japan Meteorological Agency (JMA) and the average values of weather information indicating the characteristics of the area to the input of Autoencoder, and confirmed that prediction accuracy was improved. It was also shown that prediction accuracy could be further improved by adding daytime and solar altitude data. © 2026 Institute of Electrical Engineers of Japan and Wiley Periodicals LLC.
Soke et al. (Thu,) studied this question.