基于奇异谱分析—遗传算法反向传播神经网络模型的湘江新港断面水质预测

运用奇异谱分析(SSA)对湘江新港断面557周的pH、DO、高锰酸盐指数和氨氮数据进行了预处理,再运用遗传算法优化反向传播神经网络模型进行拟合与预测。结果表明:SSA有较好的降噪能力,遗传算法反向传播(GABP)神经网络模型相比BP神经网络模型均方根误差(RMSE)平均缩小了6.96%,具有良好的预测精度;预测期内新港断面的pH、DO、高锰酸盐指数、氨氮均能满足《地表水环境质量标准》(GB 3838-2002)Ⅲ类标准,但氨氮在预测期内呈上升趋势,需警惕氨氮浓度过高引发的水污染问题。

Forecasting of water quality based on SSA-GABP neural network model in Xingang section of Xiangjiang River

Singular spectrum analysis(SSA)was used to pretreat water quality indexes of pH,DO,potassium permanganate index and ammonia nitrogen at 557 weeks of Xingang section of Xiangjiang River,and then genetic algorithm optimized back propagation neural network model was used to fit and predict the those water quality indexes.Results showed that SSA possesed a good ability for denoising.Genetic algorithm back propagation(GABP)neural network model had a better prediction accuracy compared with BP neural network,which indicated by the root mean square error(RMSE)with 6.96%reduction on average.During the forecasting period,pH,DO,potassium permanganate index and ammonia nitrogen of Xingang section all met LevelⅢof"Environmental quality standards for surface water"(GB 38385-2002).However,ammonia nitrogen would rise during the forecast period,and vigilance should be aware of water pollution problems caused by high concentrations of ammonia nitrogen.