研究目的
To propose a short-term forecasting system for solar power generation that predicts irradiance and PV power using neural network models, incorporating weather data and correction mechanisms for improved accuracy in integrating PV resources into electricity grids.
研究成果
The proposed NN-based forecasting system effectively predicts short-term irradiance and PV power with high correlation (R2 > 0.85) between forecasted and measured values. Corrections for PV module inclination and temperature improve accuracy, reducing MAPE by 2% and RMSE by 9W/m2. Future work will involve refining the system and testing in real environments to enhance its applicability for grid integration of solar power.
研究不足
The system may not accurately predict sudden changes in irradiance caused by factors like clouds. Forecast accuracy depends on the input weather data quality, and the model is tailored to specific regions, requiring adjustments for different meteorological conditions. The study uses data from a limited time period and location, which may affect generalizability.
1：Experimental Design and Method Selection:
The study employs a neural network (NN) model combined with auto-regressive integrated moving average (ARIMA) techniques for short-term irradiance and PV power forecasting. It uses historical weather data and forecasts from physical models and the Weather Research and Forecasting (WRF) model.
2：Sample Selection and Data Sources:
Historical weather data from Jeonju Weather Station in South Korea, spanning from 2010 to 2012, measured hourly. Data includes irradiance, wind speed and direction, temperature, humidity, and barometric pressure. Validation uses data from January to December
3：List of Experimental Equipment and Materials:
20 Ground weather station for observations, physical model for medium-term forecasts, WRF model for short-term forecasts, and PV modules with specified characteristics.
4：Experimental Procedures and Operational Workflow:
Input weather observations, medium-term forecasts, and WRF forecasts into the NN model. Perform irradiance forecasting every 5 minutes, apply corrections based on PV module inclination and temperature, and compute PV power outputs.
5：Data Analysis Methods:
Use mean absolute percentage error (MAPE), root mean square error (RMSE), and coefficient of determination (R2) to evaluate forecast accuracy. Analyze correlations between forecasted and measured data.
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
