研究目的
To respond the PV power ?uctuation resulting from weather change by proposing a short-term PV power forecasting algorithm using multi-layer RNN.
研究成果
The proposed multi-layer RNN-based short-term forecasting algorithm of PV power showed high accuracy in predicting PV power generation 5 minutes and 1 hour later, with accuracies of 98.02% and 93.75%, respectively. This confirms the model's applicability in responding to short-term PV power ?uctuation.
研究不足
The study focuses on short-term prediction and may require further development for long-term forecasting. Additionally, the model's performance could be affected by the quality and granularity of the IoT sensor data.
1：Experimental Design and Method Selection:
The proposed multi-layer RNN model is designed to forecast short-term PV power generation using on-site weather information collected by IoT sensors.
2：Sample Selection and Data Sources:
PV power and weather IoT data were collected from a PV power platform installed on the roof of the Engineering Building in Konkuk University, Seoul, Korea.
3：List of Experimental Equipment and Materials:
IoT sensors for measuring DC voltage and current from PV panels, and weather data such as sun radiation, module and ambient temperature, humidity, and wind speed.
4：Experimental Procedures and Operational Workflow:
The data was collected every 5 minutes, divided into training, valid, and test sets (ratio 3:1:1), and experiments were performed on the Nvidia Tesla P100 server.
5：Data Analysis Methods:
The performance was evaluated using normalized Root Mean Square Error (nRMSE).
独家科研数据包，助您复现前沿成果，加速创新突破
获取完整内容
