摘要
针对目前太阳光辐照度预测模型简单、气象要素利用不充分和预测精度较低等现象,提出一种基于RBF神经网络的太阳光辐照度预测模型。利用紫外线指数序列、地外理论辐照度序列、大气温度序列、天气类型序列和历史太阳光辐照度序列作为输入向量建立RBF神经网络预测模型,预测未来一段时间的太阳光辐照度,并采用GP-RBF算法进行网络训练。在气象台预测准确、较不准确和不准确等3种情况下分析该模型的预测情况。实验结果表明:前二种情况下太阳光辐照度的预测曲线基本与真实曲线基本吻合;在气象台预测不准确情况下预测曲线偏离不大。可见该预测模型具有实用、预测精度高等特点。
Due to the problems of meteorological elements underutilized and low prediction accuracy of current solar irradiance forecasting model, this paper presents a RBF neural network model to pre-dict solar irradiance. By using the sequences of UV index, extraterrestrial theory irradiance, air temperature and weather type solar irradiance history as the input vector sequences, a RBF neural network prediction model is set up to predict solar irradiance while using GP-RBF network training algorithm. In the three cases of meteorological forecast is accurate, less accurate and inaccurate the analysis of prediction accuracy is made. The results showed that the prediction curves of solar irradi-ance at the first two cases basically consistent with the real curves while in the case of meteorological forecasting inaccurate the deviation of prediction curve is not large. So the forecast model is practical and accurate.
出处
《广西大学学报（自然科学版）》
CAS
北大核心
2016年第5期1508-1513,共6页
Journal of Guangxi University（Natural Science Edition）
基金
国家自然科学基金资助项目(51267001)
