研究目的
To propose a simple and reliable PV forecasting method using machine learning and neural networks that utilizes only readily available weather forecasting data for accurate and robust power system security.
研究成果
The proposed PV forecasting methodology demonstrates robustness in prediction accuracy and CI effectiveness, utilizing only general weather data available to the public. It offers a practical solution for reliable power system operation, especially in areas with limited access to specialized data.
研究不足
The method's accuracy may be affected by rapid weather changes within the 3-hour forecast intervals. Additionally, the reliance on general weather data may not capture all nuances needed for highly precise predictions in all scenarios.
1：Experimental Design and Method Selection:
The methodology involves constructing multiple neural networks based on weather clustering and utilizing real-time correlation data for PV forecasting.
2：Sample Selection and Data Sources:
Uses weather data from the Japan Meteorological Agency (JMA) and solar radiation meter data from the Ministry of Economy, Trade and Industry.
3：List of Experimental Equipment and Materials:
Neural networks for forecasting, weather data, and solar radiation data.
4：Experimental Procedures and Operational Workflow:
Includes data clustering according to weather conditions, construction of neural networks, performance evaluation, and CI setting.
5：Data Analysis Methods:
Utilizes root mean square error (RMSE) and maximum error rate for performance evaluation.
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
