研究目的
Investigating the effects of various environmental parameters on the PV system output and evaluating prediction models based on Artificial Neural Networks (ANN) and regression models for selective factors to forecast solar power generation accurately.
研究成果
The ANN model can forecast the output power of the PV system with superior accuracy compared to regression models when using feature selection methods like CFS and ReliefF. This conclusion is based on the comparison of actual daily power generation graphs versus the prediction graphs of the forecasting models.
研究不足
The prediction errors on certain days were exceptionally large, possibly due to highly inaccurate weather forecasts on those days, indicating limitations in weather dependency and forecast accuracy.
1：Experimental Design and Method Selection:
The study uses regression methods and ANN-based networks for forecasting PV power output. Feature selection methods CFS and ReliefF are used to select parameters.
2：Sample Selection and Data Sources:
Data is gathered by the PV system located at the 18th Campus, Institute of Non-destructive Testing, Tomsk Polytechnic University, Tomsk, Russia.
3：List of Experimental Equipment and Materials:
The station includes a 3 kW solar battery and a 2 kW wind-driven electric plant, with specifications such as generator capacity no more than 5 kW, output voltage ~220 V, 50 Hz, height of wind generator tower 6 m, surface area of the PV modules no more than 230 m
4：Experimental Procedures and Operational Workflow:
The model consists of data pre-processing, training and testing phase, and prediction phase using the neural network model in Keras.
5：Data Analysis Methods:
The accuracy of the models is estimated by comparing the PV power forecast to the actual value, using absolute percent error calculation.
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
