研究目的
To improve system stability by providing approximated future power generation to system control engineers and facilitate dispatch of hydro power plants in an optimum way using Machine Learning (ML) algorithms for solar power forecasting.
研究成果
ML models, particularly DBN and RF, outperform the SP model in solar power forecasting under clear or overcast conditions. The study suggests further improvements by incorporating more input features and using ensemble models for different weather conditions.
研究不足
The study does not consider NWP error for longer forecasting horizons. The relationship between weather parameters and generated power may vary by region and plant design, limiting the generalizability of the models.
1：Experimental Design and Method Selection:
The study employs ML algorithms (DBN, SVMR, RF) for solar power forecasting, comparing their performance with the SP method.
2：Sample Selection and Data Sources:
Data from the Buruthakanda solar farm, including weather and generation data recorded in 2011, 2012, and 2013, is used.
3：List of Experimental Equipment and Materials:
The study uses R and R Studio for model design and development.
4：Experimental Procedures and Operational Workflow:
Data preprocessing includes normalization and filtering night-time data. ML models are trained and validated using a subset of the data.
5：Data Analysis Methods:
Performance is evaluated using RMSE, MAE, and bias metrics.
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
