Research On Intelligent Control Algorithm Of Greenhouse Environment

The process in production of the greenhouse environment has many characteristics, such as time variable, nonlinear and uncertain. It's very difficult to build an accurate mathematics model. The effect of the conventional control method (for example, PID algorithm and optimal control) is not satisfying, so intelligent control method is proposed for controlling the greenhouse environment.Fuzzy Neural Networks Controller (FNNC) is designed in this paper, which combines the advantages of neural networks and fuzzy logic to improve the learning and controlling performance of the whole system. Fuzzy Neural Networks Controller not only can process fuzzy information and finish reasoning function, but also has some neural network characteristics, such as nonlinear, self-learning, distributed memory, parallel processing. Fuzzy Neural Networks can be viewed as a parallel and distributed network, in which each neuron represents parameter in fuzzy system and each output of neuron is connected to the neuron in the next layer through the weight parameter. The fuzzy neural networks are trained continuously by inputting specimen data; then the membership function parameters and the weights of fuzzy logic rules are optimized by using back propagation algorithm. The parallel processing network makes the self-adaptation of the membership functions and the self-organization fuzzy logic rules possible. The experiments demonstrated that learning-rate and smoothness-factor can affect the convergence and convergent speed of the neural networks. Finally, the optimal learning-rate and smoothness-factor are chosen in this paper.In addition, the greenhouse environment temperature model is built by using artificial neural networks method. The network models are trained by inputting specimen data .The experiments results demonstrated the network models with different input-output time delay parameters have different precision. Finally, the most accurate model is adopted in this paper.The simulation of using FNNC to control the greenhouse environment is carried out. Comparison with PID control and fuzzy control, the simulation results showed FNNC has a better robustness and a stronger performance of tracing. In short, the Fuzzy Neural Networks method is very feasible for the greenhouse environment control.