基于概率神经网络的指标规范值的水质评价模型

MODEL OF WATER QUALITY EVALUATION WITH NORMALIZED INDICES VALUES BASED ON PROBALISTIC NEURAL NETWORKS

为了建立地表水、地下水和富营养化3类水体共同适用的概率神经网络的水质评价模型,适当设定3类水体各项指标参照值及指标值的规范变换式,使3类水体不同指标的同级标准规范值差异尽可能小,从而各指标都可用同一个"等效"规范指标替代,因此,概率神经网络隐层各类模式的基函数中心矢量的各指标分量值可视作是相同的,即都等于同级标准所有72项指标规范值的均值。基于概率神经网络的指标规范值表示的水质评价模型用于3类水体评价实例检验,验证了该模型的普适性、简洁性和实用性。

In order to estalish the model of water quality evaluation based on probalistic neural network commonly used for surface water, groundwater and lake water body, the present work set the proper refernce values and tansformed forms for each index, and made the difference in the same grade standard values with different indices could be weakened. Furthermore, the normalized values of different indices were equivalent to a certain normalized index, therefore, each index component value for center vector of basis function on each mode, which was equivalient to the mean of normalized values of 72 item indices in the same grade standard, was taken as the same namely, universal and simple. The NV-PNN model for water quality evaluation with normalized indices valuees was made use of some cases analysis, the results show that the NV-PNN model exhibits the characteristics of universal, simplicty and practiclity.