研究目的
To forecast two hour ahead solar irradiance levels at a site in Northwestern Alberta, Canada using real-time solar irradiance measured both locally and at remote monitoring stations.
研究成果
The algorithm demonstrates an adequate forecasting capability and it is evident that a relatively small number of remote solar monitoring stations (as few as five) can be used to obtain peak performance from the algorithm. Providing adequate geospatial separation of the remote monitoring sites around the target site is more desirable than clustering the sites in the strictly upwind directions.
研究不足
The algorithm can only produce solar forecasts when the sun is providing non-zero solar GHI.
1：Experimental Design and Method Selection:
The study uses an Artificial Neural Network (ANN) to forecast solar irradiance levels and a genetic algorithm to determine the optimal array size and positioning of solar monitoring stations.
2：Sample Selection and Data Sources:
Half hourly solar GHI data from the National Solar Radiation Database between the years 1998 - 2015 was used.
3：List of Experimental Equipment and Materials:
MATLAB's Neural Network Toolbox and 'ga' command for implementing the genetic algorithm.
4：Experimental Procedures and Operational Workflow:
The ANN is fed real-time GHI data from the target site and from a subset of 19 possible remote solar monitoring stations. The ANN output is a single two-hours ahead solar forecast at the target site.
5：Data Analysis Methods:
The performance of the ANN was evaluated using normalized root mean squared error (nRSME), correlation coefficient, and the mean absolute percent error (MAPE).
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
