论文部分内容阅读
利用极限学习机模型解译高氯离子干扰下盐碱土中硝酸根离子选择电极响应信号,系统分析了漂移校正算法、能斯特及极限学习机模型对电极法硝态氮(NO~-_3-N)预测结果准确性的影响差异。结果表明,漂移校正算法可明显提高传感器标定方程的重复性和一致性,响应斜率及截距电位的波动范围分别缩小了3.67%和7.25%;极限学习机模型的最优隐含层节点数为14;基于极限学习机的电极法NO~-_3-N质量浓度预测模型可较好抑制盐碱土中氯离子干扰,与标准检测结果之间的最大绝对误差和均方根误差分别为6.36 mg/L和4.02 mg/L。相关研究结论可为电极法测土过程中的信号校正、数据处理模型和模型参数选取提供参考。
Using extreme learning machine model to interpret the response signal of nitrate ion selective electrode in saline-alkali soil under the interference of high chloride ion, the drift correction algorithm, Nernst and extreme learning machine model were systematically analyzed for the nitrate ion (NO ~ -3- N) Differences in the accuracy of prediction results. The results show that the drift correction algorithm can significantly improve the repeatability and consistency of the sensor calibration equation, and the fluctuation range of response slope and intercept potential are reduced by 3.67% and 7.25% respectively. The optimal hidden layer nodes of the limit learning machine model are 14, the limit of learning machine based electrode method NO ~ -_3-N mass concentration prediction model can be better inhibited the chloride ion in saline-alkali soil interference, and the standard test results between the maximum absolute error and root mean square error of 6.36 mg / L and 4.02 mg / L. The relevant research conclusions can provide reference for signal correction, data processing model and model parameter selection in the soil testing by electrode method.