应用遗传算法优化BP神经网络预测太阳能PV/T系统热电产出

Using genetic algorithm to optimize BP neural network to predict the thermo-electric output of solar PV/T system

中文关键词: PV/T  神经网络  遗传算法  仿真预测

英文关键词: PV/T  neural networks  genetic algorithms  simulation prediction

中文摘要:

      PV/T集热器系统是一种能够同时提供低品位热能和高品位电能的新型太阳能系统,在光伏发电同时回收光伏余热,降低光伏板温度的同时不仅可以提高发电效率,而且能够将余热收集起来并转化应用于供暖和生活热水系统。本文利用遗传算法优化神经网络的方法建立了太阳能光伏光热(PV/T)系统性能的仿真预测模型,并与单一(Back Propag ation) BP神经网络的预测模型进行了对比分析。仿真预测结果表明：太阳能PV/T系统性能遗传算法优化BP神经网络模型的预测值与实际值拟合度较好,且预测精确度优于单一BP神经网络模型。其中遗传BP神经网络模型预测电效率的平均相对误差为1%,相对误差小于2%的样本占比大于95%；预测蓄热水箱温度的绝对平均误差仅为0.2℃,最大相对误差不超过1%。

英文摘要:

      The PV/T collector system is a new type of solar system that can provide both low-grade thermal energy and high-grade electrical energy at the same time. It recovers photovoltaic waste heat during photovoltaic power generation, which reduces the temperature of photovoltaic panels. Temperature reduction can not only improve power generation efficiency, but also collect waste heat to make it be transformed and utilized by heating and domestic hot water systems. In this paper, an optimized neural network based on genetic algorithm is utilized to predict the performance of solar photovoltaic (PV/T) systems. And a comparative analysis is conducted between the prediction of BP neural network and GA-BP. The studied results show that GA-BP neural network model has a good agreement with the experimental result of the solar PV/T system, and the prediction accuracy is better than that used the single BP neural network model. Comparing numerical and experimental result in detail, the GA-BP neural network model indicts that the average relative error of electrical efficiency is 1%, and the proportion of samples with a relative error of less than 2% accounts for more than 95%. The absolute average error of the predicted temperature of the storage tank is only 0.2 ℃, and the maximum relative error does not exceed 1%.