多元回归和BP人工神经网络模型对混合厌氧消化产气量的预测研究

在中温且控制pH值条件下,对脂肪类单基质和城市污水厂剩余污泥进行混合厌氧消化试验。基于多元回归原理和BP人工神经网络原理,对其建立产气量预测模型。由实验数据计算得出:两个阶段多元回归模型的预测平均准确率分别为75.69%和79.29%;BP神经网络模型的预测平均准确率为79.05%。通过对比两种模型的预测结果可知,两种模型都有较高的预测准确率,但BP模型的预测准确率更高,更适用于混合厌氧消化产气量预测。

Prediction of Gas Production of Anaerobic Co-digestion Using Multiple Regression and BP Artificial Neural Network Model

At the conditions of mesophilic temperature and controlled pH value, an anaerobic digestion experiment was conducted using the mixture of fat and excess sludge from a municipal sewage treatment plant as raw material. Based on the theory of multiple regression and BP artificial neural network, models to predict the gas production rate were established. The experimental results showed that the average predictive accuracy of multiple regression model for two experimental phases was 75.69% and 79.29% respectively; meanwhile the average predictive accuracy of BP neural network model was 79. 05%. The comparison of the predictions by the two models, it showed that the predictive accuracies for both models were high, of which the BP model had higher predictive accuracy and was better for the prediction of gas production rate of the anaerobic co-digestion.