研究目的
Investigating the impact of correlation-based feature selection on the accuracy of photovoltaic (PV) power prediction.
研究成果
The study demonstrates that correlation-based feature selection improves the accuracy of PV power prediction. Specifically, selecting weather variables with a correlation coefficient greater than 0.1 reduces RMSE by 33.7%. This provides a reference guideline for selecting weather variables to maximize prediction accuracy.
研究不足
The study is limited to the analysis of eighteen weather variables and their impact on PV power prediction accuracy. The applicability of the findings may vary with different datasets or locations.
1：Experimental Design and Method Selection:
The study uses correlation-based feature selection to analyze the impact on PV power prediction accuracy. It employs Pearson correlation coefficient (PCC) for measuring linear correlation between PV power and weather variables.
2：Sample Selection and Data Sources:
The PV power dataset is measured every fifteen minutes at a specific location in Seoul, South Korea. The weather dataset consists of eighteen weather variables collected through an API provided by the Korea meteorological administration (KMA).
3：List of Experimental Equipment and Materials:
R version
4：3 is used for conducting experiments, along with the MASS package for scatter plot generation and e1071 package for support vector regression (SVR). Experimental Procedures and Operational Workflow:
The study calculates correlation coefficients, creates subsets of weather variables based on these coefficients, generates multiple prediction models using SVR, and evaluates their accuracy using root mean square error (RMSE).
5：Data Analysis Methods:
The accuracy of prediction models is compared using RMSE to identify the most accurate model.
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
