研究目的
To investigate the input data and machine learning techniques required for day-behind predictions of PV generation, within the scope of conducting informed maintenance of these systems.
研究成果
Random Forest Model 2 performed best with an average MRE of 2.70%. Irradiance, previous generation and solar position were found to be the most important variables. A 1-year training set represents the best trade-off between dataset size and modelling accuracy. Modelling accuracy is strongly correlated to weather station distance, with distances less than 5 km preferred.
研究不足
The approach cannot be applied to new or planned sites without nearby weather data, making it a necessity to source reliable weather data.
1：Experimental Design and Method Selection:
The study compares three machine learning algorithms (SVM, RF, ANN) for predicting PV power generation using historical generation and weather data.
2：Sample Selection and Data Sources:
Five years of PV generation data from four commercial building-mounted PV installations in the UK and weather data from MIDAS.
3：List of Experimental Equipment and Materials:
Not explicitly mentioned.
4：Experimental Procedures and Operational Workflow:
Data collection and processing to identify viable sites, followed by training and sensitivity analyses.
5：Data Analysis Methods:
Mean relative error (MRE) and root mean square error (RMSE) were used to evaluate algorithm performance.
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
