针对污水处理厂生化池中参数监测智能化水平不高、人力耗费较大的问题，提出基于麻雀算法-长短期记忆神经网络(Sparrow Search Algorithm-Long Short Term Memory Network，SSA-LSTM)的水质参数预测模型。以污水处理过程中好氧区溶解氧(Dissolved Oxygen，DO)、好氧区混合液悬浮固体(Mixed Liquid Suspended Solids，MLSS)质量浓度、缺氧区DO、缺氧区氧化还原电位(Oxidation-Reduction Potential，ORP)、厌氧区DO和厌氧区ORP 6个关键指标为数据样本，进行实例研究。将SSA-LSTM的预测结果与长短期记忆神经网络(Long Short-Term Memory Network，LSTM)、粒子群算法(Particle Swarm optimization-Long Short Term Memory Network，PSO-LSTM)、深度森林以及支持向量机进行对比分析，结果显示:SSA-LSTM在6个参数上的均方误差(EMSE)和决定系数(R2)均表现出更好的预测性，预测精度最高。
Real-time prediction of water quality monitoring parameters in wastewater treatment
High-precision water quality prediction can improve the intelligent level of wastewater treatment systems,which is of great significance for the real-time monitoring of parameters of sewage treatment plants.Therefore,this paper proposes a long short-term memory neural network(LSTM)prediction model,and introduces the sparrow algorithm to optimize the learning rate,training batch,iteration times,the number of neurons in the LSTM layer,and the number of neurons in the dense layer of the model,and establishes the SSA-LSTM model to improve the prediction accuracy.The monitoring data of six parameters of aerobic zone DO,aerobic zone MLSS,anoxic zone DO,anoxic zone ORP,anaerobic zone DO,and anaerobic zone ORP in the biochemical reaction tank of Nanning Ximingjiang Wastewater Treatment Plant in 2020 and 2021 were selected for example verification,and the prediction results of SSA-LSTM were compared with deep forest and support vector machine.The results show that the determination coefficients of SSA-LSTM are 0.832,0.975,0.936,0.924,0.929,and 0.887,respectively,which are 10.070％,14.571％,5.763％,20.627％,6.293％and 3.020％higher than those of deep forest,and 11.678％,59.314％,9.860％,18.159％,8.528％and 5.847％higher than those of support vector machine.The mean square errors of SSA-LSTM are 0.717,0.026,1.241,1.224,0.853 and 0.332,respectively,which are 2.710％,16.129％,8.750％,9.333％,16.373％,and 50.742％lower than those of deep forest,and 7.244％,25.714％,9.614％,10.264％,19.528％,and 60.570％lower than those of support vector machine.The prediction results of each parameter were better than those of the deep forest and support vector machine.The SSA-LSTM is compared with the LSTM optimized by the particle swarm optimization algorithm and the original LSTM.The prediction curve fitting results show that the long-term and short-term memory neural network model optimized by the sparrow algorithm has the smallest gap with the true value and the prediction accuracy is higher.The superior performance of the SSA-LSTM method can provide theoretical support for intelligent decision-making in parameter monitoring in sewage treatment plants.
万方数据
维普
