基于粒子群优化的溶解氧质量浓度支持向量回归机.docxVIP

基于粒子群优化的溶解氧质量浓度支持向量回归机 摘要 本文提出了一种基于粒子群优化的支持向量回归机（SVR）方法用于预测溶解氧质量浓度（DO）。该新方法建立在经典SVR框架之上，并采用粒子群优化算法来有效地选择最优的参数组合。训练数据来自2018年库里蒂巴省的皮道河流域的实时水质数据，包括pH、浊度、电导率以及溶解氧质量浓度（DO）。实验结果表明，基于Pso SVR的预测性能要优于传统基于遗传算法的SVR方法，平方误差系数（RMSD）和决定系数（R2）分别为0.306和0.836。预测结果也比传统的多项式线性回归（PLR）更好，RMSD和R2的值分别为0.789和0.543。因此，新的PSO SVR模型及其有效性在预测DO含量方面具有一定的应用前景。 Abstract This paper proposes a particle swarm optimization based support vector regression (SVR) approach for predicting the dissolved oxygen mass concentration (DO). The proposed new approach is built on the classical SVR framework and a particle swarm optimization algorithm is adopted to effectively select the optimal parameters combination. The training data are collected from the real-time water quality data of Pirao River basin in Koritiba Province in the year 2018, including the pH, turbidity, electrical conductivity and dissolved oxygen mass concentration (DO). The experimental results show that the prediction performance based on Pso SVR is more accurate than the traditional one based on GA-SVR, with RMSD and R2 of 0.306 and 0.836, respectively. The prediction results are also better than traditional polynomial linear regression (PLR), with RMSD and R2 of 0.789 and 0.543, respectively. Therefore, the proposed new PSO SVR model, and its effectiveness, has an application prospect in predicting DO content. Key words: Dissolved oxygen mass concentration, Support vector regression, Particle swarm optimization, Koritiba province