摘要
水质评价是多变量的非线性系统,而神经网络解决非线性问题有其独到优势,其用于水质评价是较好的方法。在评价时,神经网络参数的确定直接影响到评价的精度。通过对输入输出数据的归一化、权和阈值的初始值以及网络隐层节点数确定方法进行探讨,优化了在水质评价时BP神经网络的参数,并在实际评价中应用,从而使评价结果更客观和符合实际。
Water quality evaluation is a multivariable nonlinear system. As is known to all, neural network （NN） has its original predominance in solving the nonlinear problem, so it is a good method to be used for water quality evaluation. The parameters of the neural network have direct influence on the precision of the evaluation. Through the discussion on the normalization of input and output data, the original value of weight and threshold value, and the ascertaining method of network latent layer node numbers, the parameters of BP NN in water quality evaluation are optimized and applied in practical evaluation, thus making the evaluation outcome more objective and up to practice.
出处
《水利与建筑工程学报》
2007年第1期12-15,共4页
Journal of Water Resources and Architectural Engineering
基金
教育部高等学校博士点专项科研基金(20051117001)
关键词
神经网络
水质评价
参数确定
neural network
water quality evaluation
determination of parameter
