Simulation Of Alfalfa Solar Drying Process In Deep Bed Based On Artificial Neural Networks

In view of the great nutrition lost and bad quality of nature drying clover products, and the high drying cost of general energy and the energy wasting and polluting environment problems, etc., the solar drying experiments of alfalfa in deep bed under real time weather condition were conducted. The traditional process modeling method is not fit to the solar drying process of alfalfa in deep bed. Since the solar drying process of alfalfa in deep bed has the characters of complicated, nonlinear and random, etc., the artificial neural networks technology is used.The solar drying process of alfalfa can be divided into three processes.After analyzing the three processes, the simulation model of every process was determined. The neural networks models were established and were validated. And six conclusions were made from the study.(1) The temperature changing of the solar collector was influenced by the solar radicalization, surroundings temperature and humidity. From the regression analysis result of the solar collector temperature and the factors, it showed that the correlation coefficients of different experiments data were very different. The corresponding coefficients of the three regression equations had positive ones and negative ones. So the general equation was not able to find out.(2) The solar drying process of alfalfa in deep bed has the character of great nonlinear: the drying speeds of different parts alfalfa were not the same, and the temperatures of different locations in the same section were not the same, the temperature of the middle part alfalfa was lower than any other part temperature all the time, the drying speed of deep layer alfalfa was low. To study on the magnitude, positive or negative and the changing trend of the alfalfa surface temperature grads helped to forecast the process of drying. That the temperature difference value between exhaust air and medium and the relative humidity difference value between exhaust air and medium could show the situation of mow humidity.(3) The factors of time and the sun angle were considered in the heat characteristic simulation model of the solar air collector. The result showed that the performance of the model had improved. Temperature simulation models and the efficiency simulation models of the solar air collector were established by considering space information or not. The simulation results showed that the models could well forecast the process.(4) The time sequence factor was considered in the alfalfa surface temperature simulation model. Not only the multiformity of the network swatches was enriched, but also the network performance was improved. The different neural network simulation models for the alfalfa surface temperature were established by considering monolayer with single dimension information, two-dimensional information and three-dimensional information, and all these information of multilayer for one day and many days. The processes were precisely forecasted: EC is above 0.97 and mape is below 5.4%. The grads neural network simulation model of the alfalfa surface temperature was established,and its result for the testing sample was also quite well: EC is above 0.97 and mape is below 6%. The temperature grads simulation model of the inside layer alfalfa along the airflow direction was established. This was important to the study and management of the deep bed alfalfa drying process, and helped to forecast the process of drying in real time. This could also be referenced as optimizing the drying process.(5) The alfalfa moisture content simulation models for different experiments data of monolayer, multilayer and many days were established based on neural networks. The models imitated well and forecasted well.(6) The solar drying system simulation model of the alfalfa in Deep Bed ,including the heat characteristic simulation models of the solar air collector, alfalfa surface temperature simulation models and the alfalfa moisture content simulation models ,was established. The simulation results of the models were all well. And the real time process of the alfalfa drying could be showed well.To sum up, the simulation of solar drying model of alfalfa in Deep Bed was established based on artificial neural networks. The models could well forecast the alfalfa moisture content of different parts. The establishment of the alfalfa moisture content simulation model was very useful to the reality. This could forecast the real time alfalfa humidity and direct the real producing.