Research On Intelligence Control Greenhouse System Based On Fuzzy Neural Networks

In this thesis, the characteristics of greenhouse system are summarized, such as nonlinearity, large delay, strong coupling. A controller for greenhouse system based on neural network is designed through applying the techniques of fuzzy control, neural network, computer to this system and integrating the knowledge representation ability of fuzzy control and the self-learning and adaptive ability of neural networks. This controller combines the advantages of fuzzy logic and neural network to improve the learning ability and the performance of system. The controller based on fuzzy neural network not only deals with fuzzy information, realizes reasoning functions, but also controls the input and output of exactitude values, optimizes membership functions and weights of fuzzy logic rules through the self-learning ability of neural networks. The main research and achievements are the follows:1. The fact that how the external factors affect the temperature, humidity and how extensive executive organizations affect the temperature, humidity is discussed. Then, the controller based on fuzzy neural network for greenhouse system is designed through combining advantages of fuzzy logic and neural network. Finally, the integrated software up machine of this controller is designed and realized. Experiments show that it can achieve intelligence control for greenhouse system.2. The controller based on fuzzy neural network is designed which take into consideration six executive organizations by analyzing the circumstance needs of controlling winter greenhouse. The denotation of some executive organizations is upgraded to the form of fuzzy sets from the switch variable which can possess better effect for control.3. An improved BP learning algorithm which adds the momentum and learning rate into the weights update formula is proposed after summarizing the disadvantages of traditional BP learning algorithm.4. Simulation experiment is performed in order to validate the performance of the controller designed based on fuzzy neural network. The results show that the fuzzy neural network methods proposed in this thesis can overcome super-regulation and poor control stability which consist in conventional PID control and fuzzy control. It possesses advantages of fast reasoning, good dynamic performance and static performance, strong anti-interference capability. So it can better meet the needs of production, have a better feasibility.