Model predictive control of temperature and humidity of artificial climate chest based on neuro dynamical optimization

Abstract: Because an artificial climate chest can provide an artificial climate environment, it is widely used in biological, chemical, and agricultural science experiments. The key technology of an artificial climate chest is the accurate control of temperature and humidity. Because of a time delay character and the coupling of temperature and humidity in an artificial climate chest, it is difficult to accurately control the temperature and humidity of an artificial climate chest. Normally, Proportion Integration Differentiation (PID) and a fuzzy control method were used to control the temperature and humidity of an artificial climate chest, but the control accuracy and response speed were not satisfactory. In this paper, model predictive control (MPC) was used to control the temperature and humidity of an artificial climate chest. Online optimization is one of key problems of MPC. The temperature and humidity object of an artificial climate chest is an object with two inputs and two outputs, and the object has the characteristics of time delay and coupling. Based on the model of the temperature and humidity of an artificial climate chest, the MPC method and optimization model of an input delay object were derived. First, the temperature and humidity object of an artificial climate chest were described in state equations form. Secondly, an optimization problem in the MPC of an artificial climate chest was proposed. Thirdly, the optimization problem of the input delay object was transformed and depicted as a quadratic programming problem (QP). Then the neurodynamical optimization was used as an online optimizer of MPC and a MPC method based on neurodynamical optimization was obtained. The neurodynamical optimization is an optimization method that incorporates an artificial neural network and dynamical system technology. Because of the inherent nature of parallel and distributed information processing in neural networks, the convergence rate of the solution process was not decreasing as the size of the problem increased. Moreover, neural networks can be implemented physically in designated hardware such as Application Specific Integrated Circuits (ASICs) where optimization is carried out in a truly parallel and distributed manner. So, neural networks are widely used in dynamical optimization problems. In section four, simulation experiments were taken using the proposed MPC method based on the neurodynamical optimization method and the PID control method. Simulation results showed that this method had smaller overshoot and higher control accuracy than the PID method. Moreover, this method can be used in other linear and nonlinear systems.