多元回归和BP人工神经网络在预测混合厌氧消化产气量过程中的应用比较

对脂肪类单基质和城市污水厂剩余污泥混合厌氧消化过程的产气阶段进行基于多元回归和BP人工神经网络的产气量预测模型比较研究。实验数据分别取自反应过程的第1～16天和第17～70天。结果表明: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2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。

Comparing multiple regression and BP artificial nerve net model used on prediction of anaerobic co-digestion gas-producing process

A comparative study on the forecasted gas-producing model based on the multiple regression and BP artificial nerve net of the gas-producing phase on an anaerobic co-digestion experiment with the fat biomass and sewage sludge was carried out. The data of the experiment was taken during the reaction process in 1^th~16 ^th and 17 ^th~70 ^th. The results showed that the average forecast correctness rate of multiple regression model was about 75.69% and 79.29%, respectively and that of BP neural network model was about 79.05%. The forecasted correctness rate of the both was higher by comparing the predicted results of the both models. However, the BP model was better than another one, which was more suitable for the gas prediction of the co-digestion system.