SOFT MEASUREMENT PREDICTION MODEL OF BIOGAS PRODUCTION BAESD ON IMPROVED PARTICLE SWARM OPTIMIZATION BP NEURAL NETWORK
The high-temperature anaerobic fermentation system of cow dung involves complex dynamics,with serious nonlinearity and time-varying,which makes it difficult to construct an accurate prediction model of biogas production.In order to accurately predict the daily gas production of large and medium-sized biogas projects,a new method of using improved particle swarm optimization algorithm to optimize traditional BP neural network was proposed,and an improved PSO-BP model was established.Firstly,the parameters are selected by soft measurement technology.Secondly,the model was established by taking the parameters such as feed rate,fermentation temperature,liquid level and hydraulic pressure in the tank as input and the daily biogas production as output.On this basis,the particle swarm optimization algorithm is improved by using the linear reduction weight coefficient method and introducing the mutation operator,and the BP neural network is initialized to improve the performance of the model.The performance of the improved PSO-BP model,the traditional BP neural network and the BP neural network optimized by genetic algorithm in predicting the daily biogas production was compared through experiments.When the improved PSO-BP model was used for prediction,the root mean square error(RMSE),the coefficient of determination(R2)and the mean absolute error(MAE)were 1.38440,0.84011 and 1.00910,respectively.It is proved that the improved PSO-BP model combined with soft measurement technology is feasible to predict the complex nonlinear high temperature anaerobic fermentation process of cow dung,and the accuracy of the prediction results is ensured.
NETL
NSTL
万方数据
